Browse source

Changelog

Query
From:
To:
Branch:
Path:
Message:
BugID:
User: +
Format:
: 2 checkins (+14/-0) by Henrik Grubbström (Grubba) <grubba@grubba.org> : 33 checkins (+2583/-3172) by 2 people : 38 checkins (+297/-177) by 3 people : 13 checkins (+89/-38) by 3 people : 64 checkins (+1191/-783) by 5 people : 32 checkins (+330/-148) by 4 people : 14 checkins (+268/-162) by 2 people : 38 checkins (+467/-710) by 3 people : 32 checkins (+1433/-564) by 2 people : 23 checkins (+376/-270) by 4 people : 38 checkins (+790/-714) by 2 people : 12 checkins (+118/-61) by 4 people : 41 checkins (+460/-342) by 4 people : 9 checkins (+353/-204) by 4 people : 10 checkins (+155/-172) by 2 people : 7 checkins (+37/-14) by 2 people : 64 checkins (+926/-491) by 4 people : 43 checkins (+257/-235) by 4 people : 22 checkins (+183/-82) by 4 people : 8 checkins (+56/-16) by 2 people : 19 checkins (+221/-99) by 4 people : 22 checkins (+229/-61) by 4 people : 5 checkins (+12/-7) by Henrik Grubbström (Grubba) <grubba@grubba.org> : 21 checkins (+121/-36) by 3 people : 19 checkins (+282/-71) by 6 people : 13 checkins (+287/-87) by 2 people : 6 checkins (+106/-79) by 2 people : 17 checkins (+1452/-143) by 3 people : 130 checkins (+859/-486) by 6 people : 36 checkins (+1379/-127) by 3 people : 22 checkins (+159/-199) by 4 people : 6 checkins (+38/-12) by Henrik Grubbström (Grubba) <grubba@grubba.org> : 1 checkin (+13/-3) by Henrik Grubbström (Grubba) <grubba@grubba.org> : 3 checkins (+172/-6) by 2 people : 4 checkins (+22/-2) by 2 people : 6 checkins (+157/-97) by 2 people : 9 checkins (+120/-100) by 2 people : 7 checkins (+102/-16) by Henrik Grubbström (Grubba) <grubba@grubba.org> : 7 checkins (+57/-7) by 3 people : 2 checkins (+13/-5) by Henrik Grubbström (Grubba) <grubba@grubba.org> : 3 checkins (+32/-2) by 2 people Lines added/removed recently

Yesterday; Thursday 18 December 2014

2014-12-18 13:24:32 (14 hours ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

Gmp: Disable Gmp.smpz()->invert().

This function has multiple issues:

* It doesn't work as implemented (eg argument 6 to mpb_sec_invert()
is wrong).

* It would clobber THIS. Gmp manual 8.1:
"In either case, the input A is destroyed."

* To work, the number of limbs in THIS, modulo and res MUST
be the same (aka n). This can probably be accomplished
by using mpz_realloc2(), of which the Gmp manual 5.1 says:
"Calling this function is never necessary; reallocation is
handled automatically by GMP when needed."

Fixing the above issues while still keeping the _sec property
is non-trivial, and best left to the Gmp people, so we wait for
an mpz_invert_sec().

2014-12-18 11:46:04 (15 hours ago) by Marcus Wellhardh <wellhardh@roxen.com>

release number bumped to 913 by export.pike
release number bumped to 912 by export.pike

Wednesday 17 December 2014

2014-12-17 17:18:42 (1 day ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

Gmp: Renaming of macros.

The old LIMBS(X) has been renamed to ALIMBS(X) (all allocated limbs).

Adds a new LIMBS(X) which returns the array of limbs.

Tuesday 16 December 2014

2014-12-16 17:47:13 (2 days ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

SSL.Cipher: validate_dh() now knows about the FFDHE groups.

Also adds the MODP groups of equivalent strength.

2014-12-16 17:00:55 (2 days ago) by Stephen R. van den Berg <srb@cuci.nl>

Fix typo.

2014-12-16 01:06:43 (3 days ago) by Bill Welliver <bill@welliver.org>

Nettle: build successfully on Windows
build: modules that invoke pike for building should now be able to do so.

Monday 15 December 2014

2014-12-15 22:10:53 (3 days ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

SSL: Support the Negotiated FF-DHE Parameters draft.

NB: This draft has been incorporated into the TLS 1.3 draft.

2014-12-15 20:25:19 (3 days ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

Gmp: The {,N}LIMBS() macros now actually use their argument.

2014-12-15 00:01:20 (4 days ago) by Martin Nilsson <nilsson@opera.com>

More const. That should be all prototypes up to date.

Sunday 14 December 2014

2014-12-14 16:00:11 (4 days ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

SSL.Constants: Updated constants from FFDHE draft 4.

Negotiated FFDHE Parameters for TLS draft 4 (and 3) replaces and
removes some groups and renumbers some constants.

Also adds the FFDHE_GROUPS lookup table.

2014-12-14 15:50:54 (4 days ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

Crypto.DH: Renamed the DLDHE constants to FFDHE for clairity.

The DLDHE draft has been renamed FFDHE. As these constants aren't yet
used anywhere, rename them accordingly.

Also updates the references for the constants to the latest drafts
where they are defined.
Crypto.DH: Renamed the DLDHE constants to FFDHE for clairity.

The DLDHE draft has been renamed FFDHE. As these constants aren't yet
used anywhere, rename them accordingly.

Also updates the references for the constants to the latest drafts
where they are defined.

2014-12-14 15:46:31 (4 days ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

Crypto.DH: Renamed the DLDHE constants to FFDHE for clairity.

The DLDHE draft has been renamed FFDHE. As these constants aren't yet
used anywhere, rename them accordingly.

Also updates the references for the constants to the latest drafts
where they are defined.

Saturday 13 December 2014

2014-12-13 22:59:15 (5 days ago) by Martin Nilsson <nilsson@opera.com>

These catches are intended to catch buffer underflow. That is however already done in the caller in Connection.

2014-12-13 13:11:52 (6 days ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

SSL [DHE]: Validate the Ys received from the peer.

This is required by the current FFDHE draft, and is needed to protect
the connection from MITM attacks.

Also changes some alerts from unexpected_message to handshake_failure
as required by the same draft.

2014-12-13 12:58:48 (6 days ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

SSL [DHE]: Validate the Ys received from the peer.

This is required by the current FFDHE draft, and is needed to protect
the connection from MITM attacks.

Also changes some alerts from unexpected_message to handshake_failure
as required by the same draft.

Friday 12 December 2014

2014-12-12 17:10:24 (6 days ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

Testsuite: Preparations for testing TLS 1.3.

The SSL testsuite now knows some about which suites that are to be
obsoleted in TLS 1.3.

2014-12-12 10:25:56 (7 days ago) by Per Cederqvist <cederp@opera.com>

Handle read errors in Protocols.LysKOM.Raw.

If there is a network outage, con->read() will return 0 (not the empty
string) after a (long) while. Treat that properly by calling
connection_lost() in that case as well.

Fixed a backrace in conwrite by returning 0 if the connection
is closed.

(cherry picked from commit 7cdea47d8c28017dbdbff49607c811ca0472000c)
Handle read errors in Protocols.LysKOM.Raw.

If there is a network outage, con->read() will return 0 (not the empty
string) after a (long) while. Treat that properly by calling
connection_lost() in that case as well.

Fixed a backrace in conwrite by returning 0 if the connection
is closed.

(cherry picked from commit 7cdea47d8c28017dbdbff49607c811ca0472000c)

2014-12-12 10:23:54 (7 days ago) by Per Cederqvist <cederp@opera.com>

Handle read errors in Protocols.LysKOM.Raw.

If there is a network outage, con->read() will return 0 (not the empty
string) after a (long) while. Treat that properly by calling
connection_lost() in that case as well.

Fixed a backrace in conwrite by returning 0 if the connection
is closed.

Thursday 11 December 2014

2014-12-11 18:52:53 (1 week ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

Nettle.ECC: Added initial implementation of Curve.Point.

This is to simplify handling of points on ECC curves.

Currently the code is sufficient to perform ECDSA, but
eg encode and decode would be nice.

TODO: Other code needs to be updated to accept Points.

2014-12-11 11:23:48 (1 week ago) by Stephen R. van den Berg <srb@cuci.nl>

pgsql: Explicitly dislodge read mutex upon fd close.

2014-12-11 11:16:54 (1 week ago) by Stephen R. van den Berg <srb@cuci.nl>

pgsql: Explicitly dislodge read mutex upon fd close.

Wednesday 10 December 2014

2014-12-10 13:39:15 (1 week ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

SSL.ServerConnection: Stricter TLS 1.3 interpretation.

Tuesday 09 December 2014

2014-12-09 18:28:55 (1 week ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

Compiler: Fixed some NULL-dereferences.

expr4 nonterminals can be NULL for eg undeclared variables.

Thanks to Jakub Krajewski <jkrajewski@opera.com> for the report.

Fixes [Pike mailinglist 14323].
Compiler: Fixed some NULL-dereferences.

expr4 nonterminals can be NULL for eg undeclared variables.

Thanks to Jakub Krajewski <jkrajewski@opera.com> for the report.

Fixes [Pike mailinglist 14323].

2014-12-09 17:26:04 (1 week ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

Master: Complain about inherited joinnodes and dirnodes.

A common cause for this is an attempt to inherit a module that is a dirnode.

Complaining at creation time makes it easier to find the mistake.

2014-12-09 17:23:14 (1 week ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

Master: Complain about inherited joinnodes and dirnodes.

A common cause for this is an attempt to inherit a module that is a dirnode.

Complaining at creation time makes it easier to find the mistake.

Monday 08 December 2014

2014-12-08 18:33:44 (1 week ago) by Jonas Walldén <jonasw@roxen.com>

Fixed the fix.

2014-12-08 17:29:25 (1 week ago) by Martin Nilsson <nilsson@opera.com>

Simplified
pike_memory_locks isn't used. Remove that and the now always unused initializer.

2014-12-08 17:23:08 (1 week ago) by Martin Nilsson <nilsson@opera.com>

Move strdup from pike_memory to port.

2014-12-08 16:35:18 (1 week ago) by Martin Nilsson <nilsson@opera.com>

Fixed decrypt_char

2014-12-08 10:57:24 (2 weeks ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

Build: Yet another strdup() fix...
Compiler: Fixed NULL-deref in tree optimizer.

CAR for a F_*_EQ node may be NULL if the variable hasn't been declared.

Fixes [bug 7397].

2014-12-08 10:53:03 (2 weeks ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

Compiler: Fixed NULL-deref in tree optimizer.

CAR for a F_*_EQ node may be NULL if the variable hasn't been declared.

Fixes [bug 7397].

2014-12-08 10:17:13 (2 weeks ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

Build: Added a missing const.

Building --with-debug should now work again.

2014-12-08 01:22:41 (2 weeks ago) by Martin Nilsson <nilsson@opera.com>

More const. Removed locale dependent cmp.

2014-12-08 01:19:39 (2 weeks ago) by Martin Nilsson <nilsson@opera.com>

Grubba wants these back.

2014-12-08 01:18:00 (2 weeks ago) by Bill Welliver <bill@welliver.org>

windows: need stdint.h for uint*

Sunday 07 December 2014

2014-12-07 20:40:14 (2 weeks ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

Windows: Improved robustness of strdup() workaround.

2014-12-07 20:35:11 (2 weeks ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

Windows: Fixed memory leak.

These malloc's were added in an attempt to track down a bug,
and shouldn't have been in the previous commit.

Also restores the mode bits on srs/main.c.

2014-12-07 19:44:05 (2 weeks ago) by Bill Welliver <bill@welliver.org>

windows: calculate trailing/leading zeros (more) correctly.
windows: strdup is deprecated.

2014-12-07 19:26:47 (2 weeks ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

Compat: Made some symbols in 8.0::Standards.ASN1.Types optional.

These symbols are not present in predef::Standards.ASN1.Types, and under
some circumstanses such objects may get passed to compat functions.

Fixes testsuite for 8.0 SSL compat where Standards.PKCS.parse_private_key()
was passed a predef::Standards.ASN1.Types.Sequence from
8.0::Standards.ASN1.Decode.simple_der_decode().

2014-12-07 17:32:17 (2 weeks ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

Threads: Propagate thread abortion errors to wait().

A thread terminating by throwing an error now causes any corresponding
wait() on it to rethrow the error (instead of just returning zero (0)).

This simplifies detection of thread failures.
Runtime: Renumbered the types for lvalue types.

Stack types for values that aren't reference counted
MUST NOT have bit 3 (8) set.

Fixes type corruption bug for amd64 and T_SVALUE_PTR (and
probably some others).

2014-12-07 17:22:21 (2 weeks ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

Threads: Propagate thread abortion errors to wait().

A thread terminating by throwing an error now causes any corresponding
wait() on it to rethrow the error (instead of just returning zero (0)).

This simplifies detection of thread failures.

2014-12-07 17:17:03 (2 weeks ago) by Bill Welliver <bill@welliver.org>

build: fix errors on Windows
build: more deattributification for Windows

2014-12-07 17:01:55 (2 weeks ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

Build: Fixed building on non-WIN32.

2014-12-07 16:40:05 (2 weeks ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

Runtime: Renumbered the types for lvalue types.

Stack types for values that aren't reference counted
MUST NOT have bit 3 (8) set.

Fixes type corruption bug for amd64 and T_SVALUE_PTR (and
probably some others).
Compat: Added 8.0::Standards.PKCS.RSA.

2014-12-07 16:29:16 (2 weeks ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

Compat: Added 8.0::Standards.PKCS.RSA.

2014-12-07 15:54:16 (2 weeks ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

Tools.AutoDoc.BMMLParser: Fixed warnings.

2014-12-07 15:53:00 (2 weeks ago) by Bill Welliver <bill@welliver.org>

build: fix uneven ifdefs

2014-12-07 15:44:58 (2 weeks ago) by Tobias S. Josefowitz <tobij@tobij.de>

Skeleton implementation of AARCH32 (ARM) machine code support

To be extended, also into AARCH64 machine code generation

Author: Arne Goedeke <el@laramies.com>
Author: Tobias S. Josefowitz <tobij@tobij.de>

2014-12-07 15:43:31 (2 weeks ago) by Bill Welliver <bill@welliver.org>

build: remove check for ssize_t and add check for io.h

2014-12-07 15:42:21 (2 weeks ago) by Tobias S. Josefowitz <tobij@tobij.de>

Nettle: be a bit better at providing pike_nettle_hash_update_func

2014-12-07 01:08:22 (2 weeks ago) by Tobias S. Josefowitz <tobij@tobij.de>

[compilefix] Gmp: fix typo

Saturday 06 December 2014

2014-12-06 23:34:20 (2 weeks ago) by Bill Welliver <bill@welliver.org>

build: change visibility for Windows build
build: attempting to work around Windows unhappiness. May not stand.

2014-12-06 21:59:14 (2 weeks ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

Stdio.Buffer: Use ptrdiff_t instead of ssize_t.

ptrdiff_t is C89, while ssize_t is POSIX. As we support some
non-POSIX architectures, use the more portable type.

2014-12-06 21:56:40 (2 weeks ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

Stdio.Buffer: Use ptrdiff_t instead of ssize_t.

ptrdiff_t is C89, while ssize_t is POSIX. As we support some
non-POSIX architectures, use the more portable type.

2014-12-06 21:51:50 (2 weeks ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

Revert "Removed some (incorrect) dependencies."

These dependencies are required to get the generated files
in the correct place.

This reverts commit 51f78b2bfb92806221f87249b419b00769ec976d.

2014-12-06 21:48:37 (2 weeks ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

Revert "Ignore UnixWare and remove some trivial dependencies."

These dependencies are required to get the generated C-files
in the correct place.

This reverts commit 38e148fd9787acae3ee07c93331a976eb566cbd0.

2014-12-06 21:45:43 (2 weeks ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

Build: Make clean should not remove lexer.h...

2014-12-06 20:14:32 (2 weeks ago) by Bill Welliver <bill@welliver.org>

build: more windows fixes
build: complete removal of --enable-dll option in order to build
successfully on Windows.
build: get COM working on Windows.
build: executable extension is not required (nor desireable) when building under Windows.
build: Windows buildability fixes.
build: starting to get CritBit working on windows.
build: help build on Windows
build: more fixes for windows.

2014-12-06 19:51:43 (2 weeks ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

Build: Fixed some warnings in Parser.HTML.

2014-12-06 19:36:04 (2 weeks ago) by Per Hedbor <ph@opera.com>

Minimal optimization of mov_imm_reg

2014-12-06 19:07:39 (2 weeks ago) by Per Hedbor <ph@opera.com>

Minimal optimization of mov_imm_reg

2014-12-06 19:05:48 (2 weeks ago) by Marcus Comstedt <marcus@mc.pp.se>

Stdio.Buffer: Fix alignment problem in io_append_{short,int}_uc

2014-12-06 18:46:29 (2 weeks ago) by Martin Nilsson <nilsson@opera.com>

Removed some (incorrect) dependencies.

2014-12-06 18:27:53 (2 weeks ago) by Martin Nilsson <nilsson@opera.com>

Ignore UnixWare and remove some trivial dependencies.

2014-12-06 18:25:55 (2 weeks ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

Compiler [amd64]: Fixed F_*CALL_BUILTIN* --with-debug.

The use of ins_debug_instr_prologue() zapped ARG1_REG for at least
F_MARK_CALL_BUILTIN.
Compiler [amd64]: Fixed F_*CALL_BUILTIN* --with-debug.

The use of ins_debug_instr_prologue() zapped ARG1_REG for at least
F_MARK_CALL_BUILTIN.

2014-12-06 17:42:21 (2 weeks ago) by Martin Nilsson <nilsson@opera.com>

rekill

2014-12-06 17:39:41 (2 weeks ago) by Per Hedbor <ph@opera.com>

Stop re-generating the prime table each build

Since the list of the first X primes rather seldom change
in this universe, keep it as a constant file instead.

Conflicts:
src/modules/Gmp/Makefile.in
src/modules/Gmp/next_prime.c

2014-12-06 17:31:50 (2 weeks ago) by Marcus Comstedt <marcus@mc.pp.se>

NetUtils: Ignore interfaces which are not up
Calendar: Increase fuzz in deltat test

2014-12-06 17:28:26 (2 weeks ago) by Martin Nilsson <nilsson@opera.com>

Stop generating lexer.h clones.

2014-12-06 17:26:25 (2 weeks ago) by Marcus Comstedt <marcus@mc.pp.se>

NetUtils: Ignore interfaces which are not up

2014-12-06 17:18:04 (2 weeks ago) by Martin Nilsson <nilsson@opera.com>

Lower the number of d_flag declarations.

2014-12-06 17:14:29 (2 weeks ago) by Martin Nilsson <nilsson@opera.com>

Lower the number of d_flag declarations.

2014-12-06 16:56:08 (2 weeks ago) by Marcus Comstedt <marcus@mc.pp.se>

Testsuite: Handle endianness dependency in hash_7_0()

2014-12-06 16:55:29 (2 weeks ago) by Martin Nilsson <nilsson@opera.com>

Use the d_flag declaration in interpret.h

2014-12-06 15:57:42 (2 weeks ago) by Martin Nilsson <nilsson@opera.com>

Maybe correct warnings fix.

2014-12-06 15:54:02 (2 weeks ago) by Marcus Comstedt <marcus@mc.pp.se>

Testsuite: Handle endianness dependency in hash_7_0()

2014-12-06 15:51:42 (2 weeks ago) by Marcus Comstedt <marcus@mc.pp.se>

Calendar: Increase fuzz in deltat test

2014-12-06 15:46:25 (2 weeks ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

Documentation: 7.4::hash() is no more.

Use hash_7_4() instead.

2014-12-06 13:31:09 (2 weeks ago) by Per Hedbor <ph@opera.com>

Stop re-generating the prime table each build

Since the list of the first X primes rather seldom change
in this universe, keep it as a constant file instead.

2014-12-06 13:08:15 (2 weeks ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

Build: Fixed warning in version.c.
Build: Fixed warning in version.c.

2014-12-06 12:19:51 (2 weeks ago) by Arne Goedeke <el@laramies.com>

Testsuite: remove tests that expect random_seed to work
Testsuite: remove tests that expect random_seed to work

2014-12-06 11:35:44 (2 weeks ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

Crypto.DH.DLDHE2048: Now with the correct value...
Crypto.DH.DLDHE2048: Now with the correct value...

2014-12-06 11:30:55 (2 weeks ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

Crypto.DH: Added DLDHE2048 from the latest FF-DHE draft.

2014-12-06 11:28:46 (2 weeks ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

Crypto.DH: Added DLDHE2048 from the latest FF-DHE draft.

2014-12-06 11:16:35 (2 weeks ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

Testsuite: Added some tests of client certificates.

Friday 05 December 2014

2014-12-05 16:40:48 (2 weeks ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

Compiler [amd64]: Fixed code generator for INC/DEC.

Fixes [bug 7384].
Compiler [amd64]: Fixed code generator for INC/DEC.

Fixes [bug 7384].

2014-12-05 16:13:11 (2 weeks ago) by Martin Nilsson <nilsson@opera.com>

Optimizations (mostly).
Optimizations (mostly).

2014-12-05 16:01:31 (2 weeks ago) by Martin Nilsson <nilsson@opera.com>

check_type_string is a noop. Remove it.
Fixed unused argument warning.

2014-12-05 14:21:34 (2 weeks ago) by Bill Welliver <bill@welliver.org>

Filesystem.Zip: New filesystem class with support for operation on Zip files.
master: add support for searching ZIP archives as part of the module path.

2014-12-05 12:52:23 (2 weeks ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

Testsuite: Added some tests of [bug 7384].

2014-12-05 09:44:34 (2 weeks ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

SSL.Connection: Received fatal alerts now invalidate the session.

2014-12-05 09:42:24 (2 weeks ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

SSL.Connection: Received fatal alerts now invalidate the session.

Thursday 04 December 2014

2014-12-04 18:40:14 (2 weeks ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

SSL.Cipher: Minor code cleanup.

2014-12-04 18:03:49 (2 weeks ago) by Martin Nilsson <nilsson@opera.com>

Move some defines into the header file.

2014-12-04 17:09:19 (2 weeks ago) by Martin Nilsson <nilsson@opera.com>

Clean up and actually use variant here. Produces better code and improved types.

2014-12-04 16:58:39 (2 weeks ago) by Martin Nilsson <nilsson@opera.com>

Updated the fallback implementation for object hashing.

Wednesday 03 December 2014

2014-12-03 19:47:48 (2 weeks ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

SSL: Reordered some constants for clarity.

Also removes some redundant stuff from the testsuite.

2014-12-03 18:29:29 (2 weeks ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

SSL: TLS 1.3 draft 3 does not allow compression.

2014-12-03 17:58:35 (2 weeks ago) by Martin Nilsson <nilsson@opera.com>

A few more error cases.

2014-12-03 17:53:37 (2 weeks ago) by Martin Nilsson <nilsson@opera.com>

Got rid of fp and sp macros. Define Pike_fp locally in functions.
Some smpz tests.
Test the hash-a-buffer features.

2014-12-03 17:51:31 (2 weeks ago) by Martin Nilsson <nilsson@opera.com>

Don't crash on lazy initialized string buffers.
Don't crash on lazy initialized string buffers.

2014-12-03 15:50:19 (2 weeks ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

Calendar: Fixed deltat calculation and tables.

Also adds some tests.
Calendar: Fixed deltat calculation and tables.

Also adds some tests.

Tuesday 02 December 2014

2014-12-02 20:43:59 (2 weeks ago) by Arne Goedeke <el@laramies.com>

Configure: fixed test for __attribute__ support

va_start and other macros expect the va_list as first argument.
This test only produced warnings with most compilers, which is
why this problem was never fixed.

2014-12-02 20:40:55 (2 weeks ago) by Arne Goedeke <el@laramies.com>

Configure: fixed test for __attribute__ support

va_start and other macros expect the va_list as first argument.
This test only produced warnings with most compilers, which is
why this problem was never fixed.

2014-12-02 19:56:01 (2 weeks ago) by Martin Nilsson <nilsson@opera.com>

Eliminated a lot of Pike_fp. Saves a few percent of binary size. Should be somewhat faster too.

2014-12-02 19:24:24 (2 weeks ago) by Stephen R. van den Berg <srb@cuci.nl>

pgsql: Use close callback to reliably react to remote closes.
pgsql: Use close callback to reliably react to remote closes.

2014-12-02 19:06:47 (2 weeks ago) by Per Hedbor <ph@opera.com>

Added support for buffers to Crypto.X.hash()

It now support the various buffer objects as the data source argument.

2014-12-02 18:40:52 (2 weeks ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

SSL: Added some constants from the TLS 1.3 draft.

2014-12-02 18:10:09 (2 weeks ago) by Martin Nilsson <nilsson@opera.com>

Moved PIKE_MODULE_EXIT to last. Made a macro for Pike_fp->current_object.

2014-12-02 18:00:23 (2 weeks ago) by Martin Nilsson <nilsson@opera.com>

Fixup

2014-12-02 17:57:02 (2 weeks ago) by Martin Nilsson <nilsson@opera.com>

First attempt at a cryptographic mpz object.

2014-12-02 12:33:47 (2 weeks ago) by Martin Karlgren <marty@roxen.com>

release number bumped to 910 by export.pike
release number bumped to 911 by export.pike

Monday 01 December 2014

2014-12-01 17:06:03 (2 weeks ago) by Stephen R. van den Berg <srb@cuci.nl>

pgsql: Improve backtraces in debugging mode.

2014-12-01 17:02:50 (2 weeks ago) by Stephen R. van den Berg <srb@cuci.nl>

pgsql: Improve backtraces in debugging mode.

2014-12-01 14:46:16 (3 weeks ago) by Martin Nilsson <nilsson@opera.com>

Make cast() protected.

2014-12-01 14:23:53 (3 weeks ago) by Martin Nilsson <nilsson@opera.com>

cast tests.
Make cast() protected.
Small cleanup of find_testsuites. module_testsuite no longer supported as test file name.

2014-12-01 14:13:18 (3 weeks ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

Calendar.YMD: Improved dwim_time().

dwim_time() now knows about ISO 9601 timestamps with colons.

Fixes [roxen.com #23993].

2014-12-01 14:06:41 (3 weeks ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

Calendar.YMD: Improved dwim_time().

dwim-time() now knows about ISO 9601 timestamps with colons.

Fixes [roxen.com #23993].

2014-12-01 14:03:17 (3 weeks ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

Calendar.YMD: Improved dwim_time().

dwim-time() now knows about ISO 9601 timestamps with colons.

Fixes [roxen.com #23993].

2014-12-01 13:18:06 (3 weeks ago) by Martin Nilsson <nilsson@opera.com>

Abort processing when fatal package is sent.
The PRNG is not guaranteed to make the same output given the same seed. Remove redundant random_seed().

2014-12-01 13:06:27 (3 weeks ago) by Martin Nilsson <nilsson@opera.com>

Fix a crash.
Fix a crash.

Sunday 30 November 2014

2014-11-30 19:27:40 (3 weeks ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

SSL: Protected some create()s.

2014-11-30 19:14:41 (3 weeks ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

SSL: Protected some create()s.

Saturday 29 November 2014

2014-11-29 18:06:16 (3 weeks ago) by Arne Goedeke <el@laramies.com>

Image.PNG: handle extra IDAT data more gracefully
Image.PNG: handle extra IDAT data more gracefully

2014-11-29 17:38:31 (3 weeks ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

Testsuite: Test SSL session resumption.
SSL.ClientConnection: Use a plain Session if none is provided.

This avoids extraneous caching of client side sessions in
a cache where they will never be looked up.
SSL.Session: Default to creation with empty id.

This is useful for client-side SSL, where there typically isn't
a useful sessionid at start.

2014-11-29 17:37:19 (3 weeks ago) by Martin Nilsson <nilsson@opera.com>

OO more.
UUID v4 is essentially just an random string, so let's use random string directly instead.

2014-11-29 17:33:22 (3 weeks ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

SSL.File: Improved API for resuming sessions somewhat.
SSL.Context: Ensure that session identifiers are unique.

Now uses Standards.UUID to generate the session identifiers.

This works around a bug in the testsuite where session identifiers
apparently could be reused.

2014-11-29 17:31:04 (3 weeks ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

SSL.Context: purge_session() now works client-side too.

2014-11-29 17:30:06 (3 weeks ago) by Stephen R. van den Berg <srb@cuci.nl>

SSL.ClientConnection: Added missing constant.

2014-11-29 17:29:31 (3 weeks ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

SSL.File: Don't purge sessions on close notify failure.

When closing the connection it is common that the peer has closed the
connection directly after sending its close notification packet without
waiting for a corresponding close notification from us. This is to be
expected, and is not a reason for us to purge the session.
SSL.File: Added API for resuming sessions.
SSL.ClientConnection: Added some support for resuming sessions.

2014-11-29 17:27:07 (3 weeks ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

SSL.ClientConnection: Keep track of the session identity.

2014-11-29 13:15:29 (3 weeks ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

Testsuite: Test SSL session resumption.

2014-11-29 11:25:19 (3 weeks ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

SSL.ClientConnection: Use a plain Session if none is provided.

This avoids extraneous caching of client side sessions in
a cache where they will never be looked up.

2014-11-29 11:21:44 (3 weeks ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

SSL.Session: Default to creation with empty id.

This is useful for client-side SSL, where there typically isn't
a useful sessionid at start.

Friday 28 November 2014

2014-11-28 22:59:22 (3 weeks ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

SSL.File: Improved API for resuming sessions somewhat.

2014-11-28 20:09:14 (3 weeks ago) by Arne Goedeke <el@laramies.com>

Image.PNG: do length check before checking for IEND
Image.GIF: Fix leak

Cloning of Image objects can throw exceptions. Push the colortable to
the stack early, to prevent it from leaking in case of an error.
Image.ANY: fix use after free
Image.GIF: sanity check maxcode
Image.PNG: do length check before checking for IEND
Image.GIF: Fix leak

Cloning of Image objects can throw exceptions. Push the colortable to
the stack early, to prevent it from leaking in case of an error.
Image.ANY: fix use after free
Image.GIF: sanity check maxcode

2014-11-28 18:14:44 (3 weeks ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

Protocols.LDAP: Corrected doc for client().
Protocols.LDAP: Corrected doc for client().

2014-11-28 18:07:44 (3 weeks ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

Testsuite: Fixed some state machine bugs in the SSL testsuite.

In some cases the when both the client and server have emptied their
send buffers, the last packet of data from the client may still be in
transit, so wait for all data to arrive at the server before closing
the server side of the connection.

Fixes issue with getting spurious truncated data in the SSL testsuite.

Also fixes bug where client->connect() detecting an expected
handshaking failure caused the threaded test to backtrace.
Testsuite: Added test of Stdio.Buffer()->_sprintf().

2014-11-28 17:23:17 (3 weeks ago) by Martin Nilsson <nilsson@opera.com>

Support for decoding secret keys.

2014-11-28 17:06:55 (3 weeks ago) by Martin Nilsson <nilsson@opera.com>

Rewrote decode_public_key to use Stdio.Buffer

2014-11-28 03:34:38 (3 weeks ago) by Stephen R. van den Berg <srb@cuci.nl>

pgsql: Timeout queries only when needed.

2014-11-28 03:29:10 (3 weeks ago) by Stephen R. van den Berg <srb@cuci.nl>

pgsql: Timeout queries only when needed.

Thursday 27 November 2014

2014-11-27 23:48:11 (3 weeks ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

Testsuite: Fixed some state machine bugs in the SSL testsuite.

In some cases the when both the client and server have emptied their
send buffers, the last packet of data from the client may still be in
transit, so wait for all data to arrive at the server before closing
the server side of the connection.

Fixes issue with getting spurious truncated data in the SSL testsuite.

Also fixes bug where client->connect() detecting an expected
handshaking failure caused the threaded test to backtrace.

2014-11-27 15:56:16 (3 weeks ago) by Martin Nilsson <nilsson@opera.com>

Update comments.

2014-11-27 15:19:54 (3 weeks ago) by Martin Nilsson <nilsson@opera.com>

Use faster and better looking padding. This however limits us to 2147483648 bit RSA keys.

2014-11-27 13:40:50 (3 weeks ago) by Martin Nilsson <nilsson@opera.com>

Type fix
Abort processing when fatal package is sent.

2014-11-27 13:24:10 (3 weeks ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

Testsuite: Added test of Stdio.Buffer()->_sprintf().

Wednesday 26 November 2014

2014-11-26 22:24:58 (3 weeks ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

Stdio.Buffer: LFUN::_sprintf() must not have side effects!

sprintf("%s", buf) and sprintf("%q", buf) had the side effect
of emptying the buffer (which is not what you expect when you
use them for debug output...).

2014-11-26 22:22:02 (3 weeks ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

Stdio.Buffer: LFUN::_sprintf() must not have side effects!

sprintf("%s", buf) and sprintf("%q", buf) had the side effect
of emptying the buffer (which is not what you expect when you
use them for debug output...).

2014-11-26 17:00:07 (3 weeks ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

Compiler: Added #pragma {,no}disassemble.

When debugging the compiler it is is often useful to get
disassembly output for just a few functions in a specific
file. This pragma causes the compiler backend to dump the
result from the peep-hole optimizer to stderr for functions
defined in sections where #pragma disassemble is active.

Note that the pragma currently only has effect if pike
has been compiled --with-rtldebug (aka PIKE_DEBUG).

2014-11-26 13:48:52 (3 weeks ago) by Stephen R. van den Berg <srb@cuci.nl>

pgsql: FETCH is eligible for parallelisation too.
pgsql: FETCH is eligible for parallelisation too.

2014-11-26 13:16:18 (3 weeks ago) by Stephen R. van den Berg <srb@cuci.nl>

pgsql: Make better use of the Stdio.Buffer add/add_ints API.
pgsql: Make better use of the Stdio.Buffer add/add_ints API.

2014-11-26 13:01:42 (3 weeks ago) by Martin Nilsson <nilsson@opera.com>

Allow hash_messages to be called multiple times.

2014-11-26 12:18:29 (3 weeks ago) by Stephen R. van den Berg <srb@cuci.nl>

pgsql: Case insensitive regexps, but more readable.
pgsql: Case insensitive regexps, but more readable.

2014-11-26 11:47:31 (3 weeks ago) by Marcus Wellhardh <wellhardh@roxen.com>

release number bumped to 909 by export.pike
release number bumped to 908 by export.pike

2014-11-26 10:50:49 (3 weeks ago) by Stephen R. van den Berg <srb@cuci.nl>

pgsql: Simplify regexps, add docs.
pgsql: Simplify regexps, add docs.

2014-11-26 09:45:57 (3 weeks ago) by Stephen R. van den Berg <srb@cuci.nl>

pgsql: Make SET synchronous to avoid surprises with subsequent queries.
pgsql: Make SET synchronous to avoid surprises with subsequent queries.

Tuesday 25 November 2014

2014-11-25 17:03:38 (3 weeks ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

SSL.Connection: Interoperability fix.

The handshake messages are hashed twice, so avoid
zapping them too early.

The SSL module now interoperates with other implementations again.

FIXME: What about renegotiation?

2014-11-25 15:57:52 (3 weeks ago) by Martin Nilsson <nilsson@opera.com>

Allow buffer objects in handshake_packet() and hearbeat_packet().

2014-11-25 15:47:17 (3 weeks ago) by Martin Nilsson <nilsson@opera.com>

Narrower.

2014-11-25 15:06:48 (3 weeks ago) by Martin Nilsson <nilsson@opera.com>

Refactored send_streaming_data

2014-11-25 14:52:51 (3 weeks ago) by Martin Nilsson <nilsson@opera.com>

More OO.

2014-11-25 14:48:08 (3 weeks ago) by Martin Nilsson <nilsson@opera.com>

More OO.

2014-11-25 13:43:48 (3 weeks ago) by Stephen R. van den Berg <srb@cuci.nl>

pgsql: Using howmuch with a sizeof compare of the Buffer does not work.

howmuch is the amount we need extra. It does not relate to the amount
already in the Buffer. To prevent a race/deadlock we need extra
signaling.

2014-11-25 13:41:39 (3 weeks ago) by Stephen R. van den Berg <srb@cuci.nl>

pgsql: Using howmuch with a sizeof compare of the Buffer does not work.

howmuch is the amount we need extra. It does not relate to the amount
already in the Buffer. To prevent a race/deadlock we need extra
signaling.

2014-11-25 13:19:19 (3 weeks ago) by Martin Nilsson <nilsson@opera.com>

Renamed methods for consistency with Stdio.Buffer

2014-11-25 13:08:57 (3 weeks ago) by Stephen R. van den Berg <srb@cuci.nl>

pgsql: Cannot use add_hint() here, the types have already been set in stone.
pgsql: Cannot use add_hint() here, the types have already been set in stone.

2014-11-25 13:08:14 (3 weeks ago) by Martin Nilsson <nilsson@opera.com>

Removed bignum functions. Using them wouldn't do much.

2014-11-25 12:58:36 (3 weeks ago) by Martin Nilsson <nilsson@opera.com>

Better use of Stdio.Buffer APIs.

2014-11-25 12:48:14 (3 weeks ago) by Martin Nilsson <nilsson@opera.com>

put_fix_string() and add_data() converted to add().

2014-11-25 12:28:42 (3 weeks ago) by Martin Nilsson <nilsson@opera.com>

OO more.

2014-11-25 12:23:50 (3 weeks ago) by Martin Nilsson <nilsson@opera.com>

UUID v4 is essentially just an random string, so let's use random string directly instead.

2014-11-25 12:14:22 (3 weeks ago) by Martin Nilsson <nilsson@opera.com>

Remove handshake messages once they have been hashed. Saves 1-2k per connection.

2014-11-25 12:05:06 (3 weeks ago) by Stephen R. van den Berg <srb@cuci.nl>

pgsql: Check for the desired buffer size before returning.
pgsql: Check for the desired buffer size before returning.

2014-11-25 11:55:49 (3 weeks ago) by Stephen R. van den Berg <srb@cuci.nl>

pgsql: Avoid race during buffer fill/read.
pgsql: Avoid race during buffer fill/read.

2014-11-25 11:53:37 (3 weeks ago) by Martin Karlgren <marty@roxen.com>

release number bumped to 8.0.35 by export.pike
release number bumped to 8.0.34 by export.pike

2014-11-25 11:39:09 (3 weeks ago) by Stephen R. van den Berg <srb@cuci.nl>

pgsql: Preserve type information if PG_DEBUG is defined.
pgsql: Preserve type information if PG_DEBUG is defined.

2014-11-25 11:34:16 (3 weeks ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

Master: Removed erroneous mutex.

This mutex has survived since ~2000, and seems to never have been useful.

In Pike 8.0 and later it could cause a deadlock, due to locking order
together with the compilers internal mutex as used by decode_value():

Legend:

CM: master()->compilation_mutex (now removed).
IM: The compilers internal mutex.

Thread 1: Thread 2: Notes:
----------------------------------------------------------
decode_value()
compile() Lock IM.
low_cast_to_program()
findprog()
low_findprog()
CM->lock() Lock CM.
compile() Wait for IM.
low_cast_to_program()
findprog()
low_findprog()
CM->lock() Wait for CM.

This bug wasn't triggered before Pike 8.0 as then the compiler
forced threads to be disabled on entry.

FIXME: Do we need to retain the symbol "master()->compilation_mutex"?

Fixes [bug 7366].
Master: Removed erroneous mutex.

This mutex has survived since ~2000, and seems to never have been useful.

In Pike 8.0 and later it could cause a deadlock, due to locking order
together with the compilers internal mutex as used by decode_value():

Legend:

CM: master()->compilation_mutex (now removed).
IM: The compilers internal mutex.

Thread 1: Thread 2: Notes:
----------------------------------------------------------
decode_value()
compile() Lock IM.
low_cast_to_program()
findprog()
low_findprog()
CM->lock() Lock CM.
compile() Wait for IM.
low_cast_to_program()
findprog()
low_findprog()
CM->lock() Wait for CM.

This bug wasn't triggered before Pike 8.0 as then the compiler
forced threads to be disabled on entry.

FIXME: Do we need to retain the symbol "master()->compilation_mutex"?

Fixes [bug 7366].

2014-11-25 11:04:43 (3 weeks ago) by Stephen R. van den Berg <srb@cuci.nl>

pgsql: Slight speedup in column processing.

2014-11-25 10:59:23 (3 weeks ago) by Stephen R. van den Berg <srb@cuci.nl>

pgsql: Slight speedup in column processing.

2014-11-25 00:14:16 (3 weeks ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

SSL.Context: Ensure that session identifiers are unique.

Now uses Standards.UUID to generate the session identifiers.

This works around a bug in the testsuite where session identifiers
apparently could be reused.

Monday 24 November 2014

2014-11-24 20:02:50 (3 weeks ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

SSL.Context: purge_session() now works client-side too.

2014-11-24 19:11:35 (3 weeks ago) by Martin Nilsson <nilsson@opera.com>

Use read_hbuffer where immediately possible.

2014-11-24 18:43:14 (3 weeks ago) by Martin Nilsson <nilsson@opera.com>

Add signature_algorithms as Buffer instead of string.

2014-11-24 18:40:17 (3 weeks ago) by Martin Nilsson <nilsson@opera.com>

Mark key exchange packets as 8bit.

2014-11-24 18:31:49 (3 weeks ago) by Martin Nilsson <nilsson@opera.com>

Rewrote get_uint, put_uint, get_var_string and get_fix_string.

2014-11-24 18:22:30 (3 weeks ago) by Martin Nilsson <nilsson@opera.com>

Replaced pop_data() with read() or direct buffer usage.

2014-11-24 18:02:08 (3 weeks ago) by Martin Nilsson <nilsson@opera.com>

put_var_string -> add_hstring

2014-11-24 17:48:10 (3 weeks ago) by Martin Nilsson <nilsson@opera.com>

Deprecate ADT.struct

2014-11-24 17:29:55 (3 weeks ago) by Martin Nilsson <nilsson@opera.com>

Use SSL.Buffer instead of ADT.struct

2014-11-24 17:15:23 (3 weeks ago) by Per Hedbor <ph@opera.com>

The buffer_error_program only takes one argument.

This could probably be handled better.

2014-11-24 17:10:25 (3 weeks ago) by Per Hedbor <ph@opera.com>

The buffer_error_program only takes one argument.

This could probably be handled better.
Significantly faster encoding of integers > INT32_MAX.

This version is about 200 times faster.

2014-11-24 16:07:57 (3 weeks ago) by Martin Nilsson <nilsson@opera.com>

read_ints exsists

2014-11-24 14:13:09 (4 weeks ago) by Stephen R. van den Berg <srb@cuci.nl>

Fix some more docs.
Fix some more docs.

2014-11-24 13:59:50 (4 weeks ago) by Stephen R. van den Berg <srb@cuci.nl>

Doc fixes.
Doc fixes.

2014-11-24 13:43:28 (4 weeks ago) by Stephen R. van den Berg <srb@cuci.nl>

pgsql: Docs tuned.
pgsql: Docs tuned.

2014-11-24 13:07:57 (4 weeks ago) by Martin Nilsson <nilsson@opera.com>

Stdio.File compat test.

2014-11-24 13:05:31 (4 weeks ago) by Martin Nilsson <nilsson@opera.com>

Possibly correct sybase compat.

2014-11-24 12:59:18 (4 weeks ago) by Martin Nilsson <nilsson@opera.com>

Testcase for Parser.XML.Validating compat.

2014-11-24 12:54:29 (4 weeks ago) by Martin Nilsson <nilsson@opera.com>

Test Locale.Charset compat.

2014-11-24 12:51:57 (4 weeks ago) by Martin Nilsson <nilsson@opera.com>

Fixed Yabu compat code.

2014-11-24 12:47:32 (4 weeks ago) by Martin Nilsson <nilsson@opera.com>

Fixed Crypto compat code.

2014-11-24 11:43:46 (4 weeks ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

Compat: Added some symbols from 8.0.

2014-11-24 11:23:59 (4 weeks ago) by Stephen R. van den Berg <srb@cuci.nl>

SSL.ClientConnection: Added missing constant.

2014-11-24 10:57:26 (4 weeks ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

8.0::Locale: No need for an inherit here.

Inheriting joinnodes (or for that matter dirnodes) is not a good idea.

As compatibility modules are automatically joined with their newer
versions, there's also no need for the inherit except for documentation
purposes.

Fixes [bug 7364].

2014-11-24 10:56:28 (4 weeks ago) by Stephen R. van den Berg <srb@cuci.nl>

pgsql: /*semi*/final for efficiency.
pgsql: /*semi*/final for efficiency.

2014-11-24 10:30:48 (4 weeks ago) by Martin Nilsson <nilsson@opera.com>

Readded #pike versions outside of defined reange.
Documentation update

2014-11-24 10:11:12 (4 weeks ago) by Stephen R. van den Berg <srb@cuci.nl>

Reinstate legacy hash functions for the benefit of legacy network protocols.

Sunday 23 November 2014

2014-11-23 18:14:20 (4 weeks ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

SSL.File: Don't purge sessions on close notify failure.

When closing the connection it is common that the peer has closed the
connection directly after sending its close notification packet without
waiting for a corresponding close notification from us. This is to be
expected, and is not a reason for us to purge the session.

2014-11-23 18:10:01 (4 weeks ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

SSL.File: Added API for resuming sessions.

2014-11-23 03:30:07 (4 weeks ago) by Stephen R. van den Berg <srb@cuci.nl>

Workaround c?-> bug.

2014-11-23 03:27:47 (4 weeks ago) by Stephen R. van den Berg <srb@cuci.nl>

Workaround c?-> bug.

Saturday 22 November 2014

2014-11-22 14:10:33 (4 weeks ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

SSL.ClientConnection: Added some support for resuming sessions.

2014-11-22 11:44:07 (4 weeks ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

SSL.ClientConnection: Keep track of the session identity.

2014-11-22 03:19:51 (4 weeks ago) by Stephen R. van den Berg <srb@cuci.nl>

pgsql: Off-by-one error.
pgsql: Off-by-one error.

2014-11-22 03:10:14 (4 weeks ago) by Stephen R. van den Berg <srb@cuci.nl>

pgsql: Rebalance threads, cut away some fat and unnecessary mutexes.

The benchmark performance is now within spitting distance of the
old pgsql driver for the unbatched single queries.
For large batches (>20), the new driver beats everyone pants down.
For single batches (==1), the new driver is about 2% slower than
the old 7.8 version.
pgsql: Rebalance threads, cut away some fat and unnecessary mutexes.

The benchmark performance is now within spitting distance of the
old pgsql driver for the unbatched single queries.
For large batches (>20), the new driver beats everyone pants down.
For single batches (==1), the new driver is about 2% slower than
the old 7.8 version.

Friday 21 November 2014

2014-11-21 21:09:16 (4 weeks ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

Image.SVG: Use features in RSVG 2.14 and later.

Many of the APIs that were used in the Image.SVG module have been
deprecated by more modern APIs. Most of the relevant new APIs
were added in RSVG 2.14.

With RSVG 2.14 and later SVGs are now rendered and scaled with Cairo.

Fixes all known compilation warnings for the Image.SVG module.

2014-11-21 19:26:28 (4 weeks ago) by Stephen R. van den Berg <srb@cuci.nl>

pgsql: reconnect robustness improved.

2014-11-21 17:47:32 (4 weeks ago) by Stephen R. van den Berg <srb@cuci.nl>

pgsql: reconnect robustness improved.

2014-11-21 16:08:37 (4 weeks ago) by Per Hedbor <ph@opera.com>

Fixed speed of bignum and 64bit integer encoding in encode_value

This fixes an issue we ran into in one of our servers where encoding a
reasonably sized list of crc32 hashes (unsigned 32 bit integers, so
50% would not fit in a signed 32bit integer) took 10+ seconds.

| Pike v8.1 release 0 running Hilfe v3.5 (Incremental Pike Frontend)
| > array ints = allocate(50000,random)(1<<32);
| > array bigs = allocate(50000,random)(1<<320);
| > set format bench

Before my fix:
| > encode_value(ints),0; ... Execution: 1.24s
| > encode_value(bigs),0; ... Execution: 121.31ms

After my fix:
| > encode_value(ints),0; ... Execution: 18.58ms
| > encode_value(b),0; ... Execution: 43.85ms

So. "normal" 64 bit integers are now encoded about 66 times faster.
Actual bignums are only 3x faster.

This makes me somewhat suspicious about the speed of various bignum
functions in pike, by the way, since the only thing that should really
have been done is to remove the function call + object
creation/destruction overhead.

And it does not take even close to 1.2 seconds to create and free
50.000 bignum objects (creating and reallocating the array 'b' above
takes about 50ms).

2014-11-21 13:44:59 (4 weeks ago) by Martin Nilsson <nilsson@opera.com>

Stdio.Buffer instead of ADT.struct

2014-11-21 11:42:24 (4 weeks ago) by Martin Nilsson <nilsson@opera.com>

Changed a few ADT.struct to Stdio.Buffer

2014-11-21 10:50:51 (4 weeks ago) by Martin Nilsson <nilsson@opera.com>

Test for 8.0::ADT.struct

2014-11-21 09:46:02 (4 weeks ago) by Martin Nilsson <nilsson@opera.com>

Sort zone order.

Thursday 20 November 2014

2014-11-20 23:23:05 (4 weeks ago) by Stephen R. van den Berg <srb@cuci.nl>

pgsql: Coalesce packets to reduce overhead.
pgsql: Coalesce packets to reduce overhead.
pgsql: Only assign _portalname once, small regressionfix.

2014-11-20 23:20:44 (4 weeks ago) by Stephen R. van den Berg <srb@cuci.nl>

pgsql: Only assign _portalname once, small regressionfix.

2014-11-20 17:28:08 (4 weeks ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

Standards.URI: Support rel being absolute in combine_uri_path().

Roxen.fix_relative() calls combine_uri_path() with an empty base
and an absolute rel to get rid of path segments with "." and "..".

Fixes Roxen.fix_relative() getting a doubled initial "/".

Also adds a related FIXME in reparse_uri().
Standards.URI: Support rel being absolute in combine_uri_path().

Roxen.fix_relative() calls combine_uri_path() with an empty base
and an absolute rel to get rid of path segments with "." and "..".

Fixes Roxen.fix_relative() getting a doubled initial "/".

Also adds a related FIXME in reparse_uri().

2014-11-20 16:19:25 (4 weeks ago) by Martin Nilsson <nilsson@opera.com>

Added the aliases from RFC 5469.

2014-11-20 15:51:59 (4 weeks ago) by Martin Nilsson <nilsson@opera.com>

Even better looking.

2014-11-20 15:38:01 (4 weeks ago) by Martin Nilsson <nilsson@opera.com>

Small cleanup of point decoding using Stdio.Buffer features.
Work around precompiler type issue.

2014-11-20 14:48:21 (4 weeks ago) by Martin Nilsson <nilsson@opera.com>

Work around a compat resolver bug.

2014-11-20 13:22:04 (4 weeks ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

Build: Rerun the precompiler when it has been changed.

Adds a dependency for the generated files on Tools.Standalone.precompile.

Also rewrites the Makefile rules to not use explicit dependencies
in implicit rules, as they seem to be ignored by at least gnu-make.
Build: Rerun the precompiler when it has been changed.

Adds a dependency for the generated files on Tools.Standalone.precompile.

Also rewrites the Makefile rules to not use explicit dependencies
in implicit rules, as they seem to be ignored by at least gnu-make.

2014-11-20 13:20:30 (4 weeks ago) by Martin Nilsson <nilsson@opera.com>

Use Context.random for IV generation.
Clamp ranges on both ints and strings. Make reversed range an error.

2014-11-20 12:19:57 (4 weeks ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

precompile: Make sure to clamp the integer range correctly.

Fixes issue with Stdio.Buffer()->read_int32() having the type
function(:int(0..-1)).

Fixes [LysLysKOM 21057425]/[Pike mailinglist 14286].

NB: This bug also caused low_describe_type() to loop infinitely
when describing the return type.
precompile: Make sure to clamp the integer range correctly.

Fixes issue with Stdio.Buffer()->read_int32() having the type
function(:int(0..-1)).

Fixes [LysLysKOM 21057425]/[Pike mailinglist 14286].

NB: This bug also caused low_describe_type() to loop infinitely
when describing the return type.

2014-11-20 12:02:02 (4 weeks ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

precompile: Make sure to clamp the integer range correctly.

Fixes issue with Stdio.Buffer()->read_int32() having the type
function(:int(0..-1)).

Fixes [LysLysKOM 21057425]/[Pike mailinglist 14286].

NB: This bug also caused low_describe_type() to loop infinitely
when describing the return type.

2014-11-20 10:26:08 (4 weeks ago) by Stephen R. van den Berg <srb@cuci.nl>

Update CHANGES.

2014-11-20 10:03:12 (4 weeks ago) by Chris Angelico <rosuav@gmail.com>

Revert "Fix typo in pgsql's _processloop"

This reverts commit 22dbe66d3b3a9265038982ee3f0c56c992311f34.

OOPS! The error was mine, oops ooops ooooops.

2014-11-20 09:48:32 (4 weeks ago) by Chris Angelico <rosuav@gmail.com>

Fix typo in pgsql's _processloop

2014-11-20 09:14:42 (4 weeks ago) by Stephen R. van den Berg <srb@cuci.nl>

pgsql: Disappearing filedescriptors need to be reported later sometimes.
pgsql: Disappearing filedescriptors need to be reported later sometimes.

2014-11-20 02:49:03 (4 weeks ago) by Stephen R. van den Berg <srb@cuci.nl>

This kludges around a parser bug in Pike 8.1.

It's a bit difficult to create an isolated testcase out of this.
Somebody please look into this.

Without this patch, the following test results in the error below:

> object b=Sql.pgsql("");
/usr/local/pike/8.1.0/lib/modules/Sql.pmod/pgsql.pike:696:switch(): Conditional expression is void.
pgsql: Pull two Stdio.Buffers apart.

2014-11-20 02:46:23 (4 weeks ago) by Stephen R. van den Berg <srb@cuci.nl>

This kludges around a parser bug in Pike 8.1.

It's a bit difficult to create an isolated testcase out of this.
Somebody please look into this.

Without this patch, the following test results in the error below:

> object b=Sql.pgsql("");
/usr/local/pike/8.1.0/lib/modules/Sql.pmod/pgsql.pike:696:switch(): Conditional expression is void.

2014-11-20 02:37:32 (4 weeks ago) by Stephen R. van den Berg <srb@cuci.nl>

pgsql: Pull two Stdio.Buffers apart.

2014-11-20 00:20:01 (4 weeks ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

Build [Image.SVG]: Fixed warning.
Image.SVG: Fixed leak on error in _decode().

The options mapping could leak in serveral error conditions.

Also adjusts some related error messages.
Image.SVG: Fixed leak on error in _decode().

The options mapping could leak in serveral error conditions.

Also adjusts some related error messages.

Wednesday 19 November 2014

2014-11-19 18:18:52 (4 weeks ago) by Anders Johansson <anders@roxen.com>

Build: Support Bison 3.

2014-11-19 17:34:02 (4 weeks ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

SSL.ServerConnection: Fixed support for resuming sessions.

It's not a good idea to send the CCS packet twice...
SSL.ServerConnection: Fixed support for resuming sessions.

It's not a good idea to send the CCS packet twice...

2014-11-19 17:06:21 (4 weeks ago) by Marcus Cromnow <marcusc@opera.com>

Delay setup of backend wakeup pipe.

This delays the wakeup pipe until after the backend is actually used.

The main reason is that after fork() all copies of the process always
have the same wakeup pipe.

Waiting until the first use of the backend makes wake_up_backend()
useful again in forked processes (as long as the backend has not been
used before the fork).

Fixes a bug in turboproxy, we use a master backend system where all
modules are loaded and partially initializes in a 'master' backend,
and new backends are then forked of from the master as needed.

2014-11-19 10:33:42 (4 weeks ago) by Stephen R. van den Berg <srb@cuci.nl>

pgsql: Be less destructive, signal then clear instead.

2014-11-19 10:30:03 (4 weeks ago) by Stephen R. van den Berg <srb@cuci.nl>

pgsql: Be less destructive, signal then clear instead.

2014-11-19 00:27:26 (4 weeks ago) by Stephen R. van den Berg <srb@cuci.nl>

pgsql: Move direct multiset to a permanent pmod.
pgsql: Move direct multiset to a permanent pmod.

2014-11-19 00:02:23 (4 weeks ago) by Stephen R. van den Berg <srb@cuci.nl>

pgsql: Recouple mutexes for optimum performance.
pgsql: Recouple mutexes for optimum performance.

Tuesday 18 November 2014

2014-11-18 23:35:32 (4 weeks ago) by Stephen R. van den Berg <srb@cuci.nl>

pgsql: Remove cruft and decouple one shortmux.
pgsql: Remove cruft and decouple one shortmux.

2014-11-18 23:10:48 (4 weeks ago) by Stephen R. van den Berg <srb@cuci.nl>

pgsql: Merge shortlived mutexes to improve cache-locality.
pgsql: Merge shortlived mutexes to improve cache-locality.

2014-11-18 22:07:01 (4 weeks ago) by Stephen R. van den Berg <srb@cuci.nl>

pgsql: Eliminate parsing race for very fast servers.
pgsql: Improve debugging output.

2014-11-18 22:03:17 (4 weeks ago) by Stephen R. van den Berg <srb@cuci.nl>

pgsql: Eliminate parsing race for very fast servers.

2014-11-18 21:19:11 (4 weeks ago) by Stephen R. van den Berg <srb@cuci.nl>

pgsql: Improve debugging output.

2014-11-18 17:59:18 (4 weeks ago) by Stephen R. van den Berg <srb@cuci.nl>

pgsql: Discard data on portals that have already been closed.
pgsql: Discard data on portals that have already been closed.

2014-11-18 13:34:58 (1 month ago) by Stephen R. van den Berg <srb@cuci.nl>

pgsql: Set preparedname earlier to eliminate a race with fast servers.
pgsql: Set preparedname earlier to eliminate a race with fast servers.

2014-11-18 13:18:23 (1 month ago) by Stephen R. van den Berg <srb@cuci.nl>

pgsql: Show filedescriptor in %O.

2014-11-18 13:14:19 (1 month ago) by Stephen R. van den Berg <srb@cuci.nl>

pgsql: Show filedescriptor in %O.

2014-11-18 13:01:21 (1 month ago) by Stephen R. van den Berg <srb@cuci.nl>

pgsql: Do not bother to close unnamed statements.
pgsql: Do not bother to close unnamed statements.

2014-11-18 11:42:24 (1 month ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

Build: simple_describe_multiset() is only used in PIKE_DEBUG.
Build: simple_describe_multiset() is only used in PIKE_DEBUG.

2014-11-18 01:26:53 (1 month ago) by Stephen R. van den Berg <srb@cuci.nl>

pgsql: Move method closer to the data.

2014-11-18 01:23:53 (1 month ago) by Stephen R. van den Berg <srb@cuci.nl>

pgsql: Move method closer to the data.

Monday 17 November 2014

2014-11-17 23:51:49 (1 month ago) by Stephen R. van den Berg <srb@cuci.nl>

pgsql: Cleanup & ensure gotdatarowdesc runs at most once per query.

2014-11-17 23:49:46 (1 month ago) by Stephen R. van den Berg <srb@cuci.nl>

pgsql: Cleanup & ensure gotdatarowdesc runs at most once per query.

2014-11-17 22:32:49 (1 month ago) by Martin Karlgren <marty@roxen.com>

release number bumped to 8.0.32 by export.pike
release number bumped to 8.0.33 by export.pike

2014-11-17 22:03:31 (1 month ago) by Pontus Rodling <frigolit@frigolit.net>

Don't print errors for WSAEACCES when ports are reserved in Windows.

2014-11-17 19:02:31 (1 month ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

precompile: Allow UNDEFINED for voidable arguments.
precompile: Allow UNDEFINED for voidable arguments.
precompile: Added some newlines in the generated code.

2014-11-17 19:00:18 (1 month ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

precompile: Added some newlines in the generated code.

2014-11-17 16:50:05 (1 month ago) by Stephen R. van den Berg <srb@cuci.nl>

pgsql: Destroy condition instead of the mutex for stability.
pgsql: Destroy condition instead of the mutex for stability.

2014-11-17 16:35:46 (1 month ago) by Stephen R. van den Berg <srb@cuci.nl>

pgsql: Eliminate prepbuffer race (signal before wait and duplicate entries).

2014-11-17 16:33:02 (1 month ago) by Stephen R. van den Berg <srb@cuci.nl>

pgsql: Eliminate prepbuffer race (signal before wait and duplicate entries).

2014-11-17 12:37:39 (1 month ago) by Stephen R. van den Berg <srb@cuci.nl>

pgsql: Eliminate race while flushing output buffer.
pgsql: Eliminate race while flushing output buffer.

2014-11-17 11:05:06 (1 month ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

release number bumped to 907 by export.pike
release number bumped to 906 by export.pike

2014-11-17 06:36:12 (1 month ago) by Pontus Rodling <frigolit@frigolit.net>

Don't print errors for WSAEACCES when ports are reserved in Windows.

2014-11-17 06:24:20 (1 month ago) by Pontus Rodling <frigolit@frigolit.net>

Don't print errors for WSAEACCES when ports are reserved in Windows.

2014-11-17 02:08:56 (1 month ago) by Stephen R. van den Berg <srb@cuci.nl>

Reinstate legacy XML parsing modes until further notice.

2014-11-17 02:05:21 (1 month ago) by Stephen R. van den Berg <srb@cuci.nl>

Reinstate legacy XML parsing modes until further notice.

Sunday 16 November 2014

2014-11-16 22:35:37 (1 month ago) by Arne Goedeke <el@laramies.com>

ADT.Bloom.SValue: corrected comment

2014-11-16 22:17:48 (1 month ago) by Arne Goedeke <el@laramies.com>

ADT.Bloom.Simple: byte order independent hashing for wide strings
SipHash24: added byte order independent versions for wide strings

- the current siphash version does not know about wide strings, so
the hash value for wide strings depends on the byte order. added
versions for both 16 and 32 bit wide chars
- also replaced the manual little endian reading routines by the
get_unaligned_le64 macros, which generate much shorter code for
x86_64
- added siphash24_pike_string which will generate a byte order
independent hash value
- use siphash24_pike_string in ADT.Bloom.String

2014-11-16 14:58:15 (1 month ago) by Arne Goedeke <el@laramies.com>

ADT.Bloom.Simple: use byte order independent hashing

2014-11-16 14:35:31 (1 month ago) by Arne Goedeke <el@laramies.com>

add popcount32 and popcount64
Added ADT.Bloom

Added several different Bloom filters. They all use the same base class
but differ in what keys they accept and how the hash functions are
constructed. Consult the documentation for more information.

Bloom filters can be faster than hash tables However, this speed
advantage is hard to get to from pike, due to the extra cost of
calling lfuns vs mapping lookups. Therefore, the only advantage
these bloom filters have over pike mappings is the much lower
memory footprint.
siphash: make key and return value INT64

2014-11-16 13:37:44 (1 month ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

Build: Fixed some warnings in CritBit.
precompile: Reduce warnings in the generated code.
precompile: Reduce warnings in the generated code.

2014-11-16 13:27:38 (1 month ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

Build: Added PIKE_UNUSED_ATTRIBUTE.
Build: Added PIKE_UNUSED_ATTRIBUTE.

2014-11-16 13:14:16 (1 month ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

Build: Fixed some warnings in CritBit.

2014-11-16 13:03:00 (1 month ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

Build: Fixed warnings about unused arguments in threads.c.
Build: Fixed warnings about unused arguments in threads.c.

2014-11-16 12:59:06 (1 month ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

Manual: Updated compatibility chapter for Pike 8.1 and later.

Removes the section about Pike 7.6.

Adds a place-holder section about Pike 8.0.

2014-11-16 12:53:13 (1 month ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

Manual: Updated the compatibility chapter.

2014-11-16 12:51:11 (1 month ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

Manual: Updated the compatibility chapter.

2014-11-16 00:13:06 (1 month ago) by Martin Nilsson <nilsson@opera.com>

Add 8.0 compat and remove references to 7.0 and 7.2 compat.

2014-11-16 00:07:02 (1 month ago) by Martin Nilsson <nilsson@opera.com>

Use read_hint. It has to go through Gmp.mpz until all relevant interfaces are converted into int.
Use add_hint

Saturday 15 November 2014

2014-11-15 23:21:00 (1 month ago) by Stephen R. van den Berg <srb@cuci.nl>

Remove superfluous legacy compat support prior to 7.8.

2014-11-15 19:01:01 (1 month ago) by Stephen R. van den Berg <srb@cuci.nl>

Remove superfluous legacy compat support prior to 7.8.

2014-11-15 14:47:20 (1 month ago) by Stephen R. van den Berg <srb@cuci.nl>

Remove superfluous legacy compat support prior to 7.8.

2014-11-15 14:42:52 (1 month ago) by Stephen R. van den Berg <srb@cuci.nl>

Remove superfluous legacy compat support prior to 7.8.

2014-11-15 14:04:25 (1 month ago) by Stephen R. van den Berg <srb@cuci.nl>

Fix Stdio.File compat code.

2014-11-15 11:55:35 (1 month ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

SSL.Cipher [ecdh]: Added note about point validation.

There are attacks against ECDH/ECDHE that are based on providing
points that aren't on the negotiated curve, so add an explicit
note that the received points are validated in point_mul().
Nettle.ECC_Curve: Improved documentation about point_mul().

Added note that point_mul() validates that the coordinate
is valid for the curve.

2014-11-15 11:50:42 (1 month ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

SSL.Cipher [ecdh]: Added note about point validation.

There are attacks against ECDH/ECDHE that are based on providing
points that aren't on the negotiated curve, so add an explicit
note that the received points are validated in point_mul().

2014-11-15 11:48:08 (1 month ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

Nettle.ECC_Curve: Improved documentation about point_mul().

Added note that point_mul() validates that the coordinate
is valid for the curve.

2014-11-15 10:01:11 (1 month ago) by Stephen R. van den Berg <srb@cuci.nl>

Internal types only: enum -> macros.

2014-11-15 09:58:41 (1 month ago) by Stephen R. van den Berg <srb@cuci.nl>

Internal types only: enum -> macros.

2014-11-15 06:01:57 (1 month ago) by Martin Nilsson <nilsson@opera.com>

Compile fix. I don't have GSSAPI

2014-11-15 05:50:39 (1 month ago) by Martin Nilsson <nilsson@opera.com>

Some additional cleanup. Mostly renaming things.

2014-11-15 05:12:25 (1 month ago) by Martin Nilsson <nilsson@opera.com>

ADT.struct -> Stdio.Buffer

2014-11-15 02:21:44 (1 month ago) by Stephen R. van den Berg <srb@cuci.nl>

Update docs.
Update docs.

2014-11-15 01:59:42 (1 month ago) by Stephen R. van den Berg <srb@cuci.nl>

pgsql: Properly terminate the connection if the database does not exist.
pgsql: Properly terminate the connection if the database does not exist.

Friday 14 November 2014

2014-11-14 19:13:32 (1 month ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

Updated C-level documentation of callbacks.
Updated C-level documentation of callbacks.

2014-11-14 19:04:25 (1 month ago) by Stephen R. van den Berg <srb@cuci.nl>

pgsql: Cosmetics.

2014-11-14 18:57:05 (1 month ago) by Stephen R. van den Berg <srb@cuci.nl>

pgsql: Cosmetics.

2014-11-14 17:45:52 (1 month ago) by Stephen R. van den Berg <srb@cuci.nl>

pgsql: Release all locks early upon error.

2014-11-14 16:23:05 (1 month ago) by Stephen R. van den Berg <srb@cuci.nl>

pgsql: Release all locks early upon error.

2014-11-14 10:39:51 (1 month ago) by Stephen R. van den Berg <srb@cuci.nl>

pgsql: Connects failing due to insufficient resources handled more gracefully.

2014-11-14 10:34:29 (1 month ago) by Stephen R. van den Berg <srb@cuci.nl>

pgsql: Connects failing due to insufficient resources handled more gracefully.

2014-11-14 02:28:57 (1 month ago) by Stephen R. van den Berg <srb@cuci.nl>

pgsql: Improve backtrace clarity of PGassist.

2014-11-14 02:24:14 (1 month ago) by Stephen R. van den Berg <srb@cuci.nl>

pgsql: Improve backtrace clarity of PGassist.

2014-11-14 00:34:53 (1 month ago) by Stephen R. van den Berg <srb@cuci.nl>

pgsql: Finetuned behaviour upon initial connection failure.
pgsql: Finetuned behaviour upon initial connection failure.

Thursday 13 November 2014

2014-11-13 22:59:06 (1 month ago) by Stephen R. van den Berg <srb@cuci.nl>

pgsql: Delay accepting a commit until all portals have completed.

2014-11-13 22:52:24 (1 month ago) by Stephen R. van den Berg <srb@cuci.nl>

pgsql: Delay accepting a commit until all portals have completed.

2014-11-13 21:10:35 (1 month ago) by Arne Goedeke <el@laramies.com>

Strings: silence a warning

Using a function instead of a macro takes care of casting the first
argument to a signed type, which silences a warning when the macro
was used with unsigned types.
rename aligned_alloc to xalloc_aligned

C11 defines aligned_alloc. This can lead to all kinds of confusion, so
lets rename our internal function.
rename aligned_alloc to xalloc_aligned

C11 defines aligned_alloc. This can lead to all kinds of confusion, so
lets rename our internal function.
Strings: silence a warning

Using a function instead of a macro takes care of casting the first
argument to a signed type, which silences a warning when the macro
was used with unsigned types.

2014-11-13 16:48:22 (1 month ago) by Stephen R. van den Berg <srb@cuci.nl>

pgsql: Notify callers early when connections are not possible.

2014-11-13 16:44:24 (1 month ago) by Stephen R. van den Berg <srb@cuci.nl>

pgsql: Notify callers early when connections are not possible.

2014-11-13 15:40:58 (1 month ago) by Stephen R. van den Berg <srb@cuci.nl>

pgsql: Close race between a parse in progress and a Sync.
pgsql: Reset session urgently and deadlockfree.
pgsql: Improve debugging output and release all locks upon error.
pgsql: Reset session urgently and deadlockfree.
pgsql: Improve debugging output and release all locks upon error.
pgsql: Close race between a parse in progress and a Sync.

2014-11-13 13:50:51 (1 month ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

SSL.Connection: Fixed race condition with alert callbacks.

to_write() could get called with the pending alert before
the alert callback had determined what to do about the alert.

This caused the http to https redirect fallback in Roxen to fail,
due to to_write() having closed the stream before the callback
could steal it.

2014-11-13 13:46:58 (1 month ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

SSL.Connection: Fixed race condition with alert callbacks.

to_write() could get called with the pending alert before
the alert callback had determined what to do about the alert.

This caused the http to https redirect fallback in Roxen to fail,
due to to_write() having closed the stream before the callback
could steal it.

2014-11-13 12:24:37 (1 month ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

Process.spawn_pike: Added launcher argument.

2014-11-13 12:21:21 (1 month ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

Process.spawn_pike: Added launcher argument.

2014-11-13 11:10:55 (1 month ago) by unknown <unknown user@unknown>

release number bumped to 905 by export.pike
release number bumped to 904 by export.pike

2014-11-13 10:33:58 (1 month ago) by Stephen R. van den Berg <srb@cuci.nl>

pgsql: Reset prepared statement numbering upon cache invalidation.

2014-11-13 10:30:51 (1 month ago) by Stephen R. van den Berg <srb@cuci.nl>

pgsql: Reset prepared statement numbering upon cache invalidation.

2014-11-13 09:58:44 (1 month ago) by Stephen R. van den Berg <srb@cuci.nl>

pgsql: Slow down driver even further.

Statements eligible to be automatically preparsed and kept in the
statement cache have to be parsed synchronously too to prevent
accidental reordering of parsing order.

2014-11-13 09:55:17 (1 month ago) by Stephen R. van den Berg <srb@cuci.nl>

pgsql: Slow down driver even further.

Statements eligible to be automatically preparsed and kept in the
statement cache have to be parsed synchronously too to prevent
accidental reordering of parsing order.

2014-11-13 00:46:01 (1 month ago) by Martin Nilsson <nilsson@opera.com>

backzone has been added

2014-11-13 00:44:31 (1 month ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

Updated to tzdata2014j.
Updated to tzdata2014j.
Updated to tzdata2014j.

Wednesday 12 November 2014

2014-11-12 23:03:00 (1 month ago) by Martin Nilsson <nilsson@opera.com>

Use Stdio.Buffer instead of ADT.struct. Since ADT.struct inherit Stdio.Buffer, this is backwards compatible unless the ADT.struct class was further extended or completely replaced.
Use Stdio.Buffer instead of ADT.struct. Fixes a bug where the code was stepping back in the buffer by reducing index, which isn't supported anymore in ADT.struct.

2014-11-12 23:00:40 (1 month ago) by Martin Nilsson <nilsson@opera.com>

Use Stdio.Buffer instead of ADT.struct. Fixes a bug where the code was stepping back in the buffer by reducing index, which isn't supported anymore in ADT.struct.

2014-11-12 22:52:05 (1 month ago) by Martin Nilsson <nilsson@opera.com>

Use String.Buffer instead of ADT.struct in Hilfe test.

2014-11-12 20:46:18 (1 month ago) by Martin Karlgren <marty@roxen.com>

SNMP: Re-fix the lookup table sent to simple_der_decode.
SNMP: Re-fix the lookup table sent to simple_der_decode.

2014-11-12 17:23:32 (1 month ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

release number bumped to 903 by export.pike
release number bumped to 902 by export.pike

2014-11-12 16:19:06 (1 month ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

ZXID: Renamed internal function to avoid namespace conflicts.
ZXID: Fixed some memory leaks on Session destruct.
ZXID: Renamed internal function to avoid namespace conflicts.
ZXID: Fixed some memory leaks on Session destruct.

2014-11-12 16:04:28 (1 month ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

ZXID: Renamed internal function to avoid namespace conflicts.
ZXID: Fixed some memory leaks on Session destruct.

2014-11-12 12:10:00 (1 month ago) by Stephen R. van den Berg <srb@cuci.nl>

protected -> private where it makes sense.

2014-11-12 12:07:25 (1 month ago) by Stephen R. van den Berg <srb@cuci.nl>

protected -> private where it makes sense.

2014-11-12 11:28:31 (1 month ago) by Per Hedbor <ph@opera.com>

Fundamental properties of empty strings corrected.

An empty string:
- IS lowercase
- IS uppercase

'STRING_IS_LOWERCASE' means the same string will be returned for
lower_case(str), and 'STRING_IS_UPPERCASE' means the same string will
be returned for upper_case(str).

Having this be wrong for the empty string is not really a good idea,
since when strings are constructed using += the bits will not be set
correctly if you start with the empty string.

Also note: All code using string ranges really needs a special case
for the empty string, since its range of characters, when checked
using check_string_range, will be 0..255 (loose check), or
-MAX_INT32..MAX_INT32 (non-loose check).

This is done correctly in the 'string_has_null' function, which is
what is supposed to be used to check if strings contain null
characters.

2014-11-12 11:24:24 (1 month ago) by Per Hedbor <ph@opera.com>

Fundamental properties of empty strings corrected.

An empty string:
- IS lowercase
- IS uppercase

'STRING_IS_LOWERCASE' means the same string will be returned for
lower_case(str), and 'STRING_IS_UPPERCASE' means the same string will
be returned for upper_case(str).

Having this be wrong for the empty string is not really a good idea,
since when strings are constructed using += the bits will not be set
correctly if you start with the empty string.

Also note: All code using string ranges really needs a special case
for the empty string, since its range of characters, when checked
using check_string_range, will be 0..255 (loose check), or
-MAX_INT32..MAX_INT32 (non-loose check).

This is done correctly in the 'string_has_null' function, which is
what is supposed to be used to check if strings contain null
characters.

2014-11-12 10:42:17 (1 month ago) by Stephen R. van den Berg <srb@cuci.nl>

English grammar fix.
English grammar fix.

2014-11-12 10:39:41 (1 month ago) by Stephen R. van den Berg <srb@cuci.nl>

English grammar fix.

Tuesday 11 November 2014

2014-11-11 22:54:09 (1 month ago) by Stephen R. van den Berg <srb@cuci.nl>

Fix formatcheck to accomodate for the optional 2nd argument.

2014-11-11 22:48:48 (1 month ago) by Stephen R. van den Berg <srb@cuci.nl>

Fix formatcheck to accomodate for the optional 2nd argument.
Elaborate error message about NULs in the string.
Fundamental properties of empty strings corrected.

An empty string:
- is NOT lowercase.
- is NOT uppercase.
- does NOT have a minimum and maximum equal to zero.

2014-11-11 22:42:26 (1 month ago) by Stephen R. van den Berg <srb@cuci.nl>

Fundamental properties of empty strings corrected.

An empty string:
- is NOT lowercase.
- is NOT uppercase.
- does NOT have a minimum and maximum equal to zero.
Fix formatcheck to accomodate for the optional 2nd argument.
Elaborate error message about NULs in the string.

2014-11-11 21:19:03 (1 month ago) by Martin Karlgren <marty@roxen.com>

Standards.ASN1.Decode.simple_der_decode requires combined tag ids in "types".
Standards.ASN1.Decode.simple_der_decode requires combined tag ids in "types".

2014-11-11 20:12:59 (1 month ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

Build: Fixed some warnings about freeing static memory.
Build: Fixed some warnings about freeing static memory.

Monday 10 November 2014

2014-11-10 23:52:03 (1 month ago) by Martin Nilsson <nilsson@opera.com>

Hide GSSAPI when it isn't working.

2014-11-10 23:47:13 (1 month ago) by Stephen R. van den Berg <srb@cuci.nl>

Add documentation about the custom backend.

2014-11-10 23:42:45 (1 month ago) by Stephen R. van den Berg <srb@cuci.nl>

Add documentation about the custom backend.

2014-11-10 18:50:40 (1 month ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

Configure: Improved handling of empty ABI directories.
Configure: Improved handling of empty ABI directories.

2014-11-10 17:09:13 (1 month ago) by Stephen R. van den Berg <srb@cuci.nl>

pgsql: Reduce the number of dynamically created threads.

This patch pushes most of the dynamically created threads into the
running local_backend. It specifically does not push the gotdatarowdesc
one to avoid deadlocks.
pgsql: Reduce the number of dynamically created threads.

This patch pushes most of the dynamically created threads into the
running local_backend. It specifically does not push the gotdatarowdesc
one to avoid deadlocks.

2014-11-10 16:50:22 (1 month ago) by Stephen R. van den Berg <srb@cuci.nl>

Avoid ?-> operator for now.
Avoid ?-> operator for now.

2014-11-10 15:53:03 (1 month ago) by Stephen R. van den Berg <srb@cuci.nl>

pgsql: Run callbacks in a separate thread to keep session afloat.
pgsql: Run callbacks in a separate thread to keep session afloat.

2014-11-10 15:27:45 (1 month ago) by Stephen R. van den Berg <srb@cuci.nl>

Doc fixes.

2014-11-10 15:22:03 (1 month ago) by Stephen R. van den Berg <srb@cuci.nl>

Doc fixes.

2014-11-10 14:51:21 (1 month ago) by Stephen R. van den Berg <srb@cuci.nl>

Fix docs.
Fix docs.

2014-11-10 13:11:01 (1 month ago) by Stephen R. van den Berg <srb@cuci.nl>

Fix code to match docs.

2014-11-10 13:08:44 (1 month ago) by Stephen R. van den Berg <srb@cuci.nl>

Fix code to match docs.
pgsql: #require required.

2014-11-10 13:05:33 (1 month ago) by Stephen R. van den Berg <srb@cuci.nl>

pgsql: #require required.

2014-11-10 13:00:51 (1 month ago) by Stephen R. van den Berg <srb@cuci.nl>

pgsql: #require Thread.Thread.

2014-11-10 12:54:50 (1 month ago) by Stephen R. van den Berg <srb@cuci.nl>

pgsql: #require Thread.Thread.

2014-11-10 10:51:23 (1 month ago) by Martin Nilsson <nilsson@opera.com>

ins_short isn't used.

2014-11-10 10:50:33 (1 month ago) by Stephen R. van den Berg <srb@cuci.nl>

pgsql: Slow down driver by default, async parsing only on request.

2014-11-10 10:45:15 (1 month ago) by Stephen R. van den Berg <srb@cuci.nl>

pgsql: Slow down driver by default, async parsing only on request.

2014-11-10 02:52:29 (1 month ago) by Stephen R. van den Berg <srb@cuci.nl>

Stricter types.
Stricter types.

2014-11-10 01:41:40 (1 month ago) by Stephen R. van den Berg <srb@cuci.nl>

pgsql: Updated CHANGES.

Sunday 09 November 2014

2014-11-09 23:10:01 (1 month ago) by Stephen R. van den Berg <srb@cuci.nl>

pgsql: Make statusfunctions wait until the connection has been established.

The driver has survived moderate testing in production servers (Roxen),
it is ready for production release.
pgsql: Missed a spot where a portal could be undefined.
pgsql.resync interface fixed.
pgsql: Support Val.null.

2014-11-09 23:05:09 (1 month ago) by Stephen R. van den Berg <srb@cuci.nl>

pgsql.resync interface fixed.
pgsql: Support Val.null.
pgsql: Make statusfunctions wait until the connection has been established.

The driver has survived moderate testing in production servers (Roxen),
it is ready for production release.
pgsql: Missed a spot where a portal could be undefined.
pgsql: finetuning, cleanup, improved error response behaviour.
Sql.pgsql: New and improved full threaded version.

Lower latencies, increased parallelism, theoretical
elimination of all race conditions, simplified code,
uses Stdio.Buffer, no C-helper anymore, less filling.

As a new feature, introduces: send_row(),
fetch_row_array() for increased efficiency,
set_result_callback() and set_result_array_callback()
for event driven database queries.
pgsql: Further finetuning, fix normal query() race condition, thread safe.

Make the result object threadsafe (e.g. calling fetch_fields() and
fetch_row() simultaneously from multiple threads on the same result
object is supported).
pgsql: Support PostgreSQL 9.4 authentication quirks.
pgsql: Fixed reconnect behaviour, fixed textqueries, eliminated one race.

All issues eliminated. I will run it in production servers as of now.
Stdio.Buffer: Support all types for add_hstring.
Stdio.Buffer: Support offset parameter for add_hstring/read_hstring.

2014-11-09 20:33:48 (1 month ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

Build: Fixed warning about setjmp clobbering yyerrstatus.

2014-11-09 20:24:55 (1 month ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

Build: Fixed warning about setjmp clobbering yyerrstatus.

2014-11-09 00:13:19 (1 month ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

Build: Added configure option --enable-werror.

The configure option --enable-werror now adds -Werror
to the set of WARN flags.

2014-11-09 00:11:02 (1 month ago) by Henrik Grubbström (Grubba) <grubba@grubba.org>

Build: Added configure option --enable-werror.

The configure option --enable-werror now adds -Werror
to the set of WARN flags.

[permalink]

Bugs mentioned

  5469NEWSNMP instrumentation for call_out()
  7364RESOLVEDCompat resolver can't handle module.pmod
  7366RESOLVEDCompilation deadlock
  7384RESOLVEDBroken "--" operator on mappings
  7397RESOLVEDSegfault after for clause compilation error
  902RESOLVEDspider.XML dumps core
  903RESOLVEDPHP-4.0.5-dev gives "Internal server error".
  904RESOLVEDstealth_check_mapping_for_destruct can fail in the gc
  906RESOLVED<gtext> (with true type fonts)breaks if there is a line break in the container
  907RESOLVEDCalendar.Namedays does not compile
  908RESOLVEDcomment field has wrap=hard
  909RESOLVEDEmpty $referer in log does not set '-'.
  910RESOLVEDCreate Standard site with template doesn't work.
  911RESOLVEDHTML tags in header gtexts in RFCs.
Bugs? Suggestions?