summaryrefslogtreecommitdiffstats
Commit message (Collapse)AuthorAgeFilesLines
...
| | * | Add the critical missing UpdateInterest() call at the exit of DoRead().dgp2015-08-101-3/+10
| | | | | | | | | | | | | | | | | | | | | | | | (Compare with same approach in DoReadChars()). This involves removing some other calls that are now replaced by the new one. Also marked several UpdateInterest() calls throughout tclIO.c with comments raising the suspicion that they serve no function.
* | | | [32ae34e63a] Prevent segfaults and data corruption when CopyData() is called ↵dgp2015-09-234-25/+241
|\ \ \ \ | |/ / / | | | | | | | | | | | | recursively. Mark new test io-53.20 as "knownBug". It demos some unknown flaw in MoveBytes().
| * | | [32ae34e63a] Prevent segfaults and data corruption when CopyData() is calleddgp2015-09-234-25/+242
| |\ \ \ | | | | | | | | | | recursively.
| | * | | Update tests to account for changed ReflectWatch behavior.bug_32ae34e63adgp2015-09-231-13/+13
| | | | |
| | * | | Protect CopyState buffer from conflicting uses when CopyData() is calleddgp2015-09-232-8/+7
| | | | | | | | | | | | | | | recursively. Also, have ReflectWatch() always give driver a chance to act.
| | * | | CopyData() now tolerates recursion enough so that io-53.18 does not segfault.dgp2015-09-231-21/+9
| | | | |
| | * | | WIPdgp2015-09-231-3/+47
| | | | |
| | * | | Reboot the bug fix.dgp2015-09-221-0/+186
| | |\ \ \ | |/ / / /
| | * | | merge 8.5dgp2015-09-222-8/+10
| | |\ \ \ | | |/ / / | |/| | |
| | * | | merge 8.5dgp2015-09-023-13/+47
| | |\ \ \
| | * \ \ \ merge 8.5dgp2015-08-276-14/+33
| | |\ \ \ \
| | * | | | | New test attempting to demo Bug [e0a7b3e5f8]. Doesn't work yet.dgp2015-08-131-3/+65
| | | | | | |
| | * | | | | Disable the filtering of ReflectWatch so that every UpdateInterest() calldgp2015-08-131-1/+1
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | flows through to become a [driver watch] evaluation. This makes new test io-53.19 pass. It also makes a collection of 10 test in ioCmd.test start failing, all of which are recording detailed reflected channel driver command evaluation. The now unfiltered [driver watch] change this record without (at least apparently) changing any behavior. Need review.
| | * | | | | A bit of code safety, and then a test demonstrating next issue.dgp2015-08-132-2/+67
| | | | | | |
| | * | | | | Adjustments to failing tests to account for changed [chan postevent].dgp2015-08-121-4/+4
| | | | | | |
| | * | | | | New test io-53.18 adapted from demo script in [32ae34e63a]. This testdgp2015-08-122-1/+87
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | segfaults without changes to source code. This checkin also contains a revised implementationf of [chan postevent] that stops calling Tcl_NotifyChannel() directly, and queues an event to do it instead. This stops the segfault, but causes tests iocmd-31.[67] to fail. Need advice on whether that matters.
| | | | | | * Merge trunkjan.nijtmans2015-09-2320-645/+839
| | | | | | |\ | |_|_|_|_|_|/ |/| | | | | |
* | | | | | | Micro-optimization handling Tcl_Panic(), suggested by D. Richard Hipp.jan.nijtmans2015-09-237-16/+31
|\ \ \ \ \ \ \ | | | | | | | | | | | | | | | | As a side-effect, this is as well a fix for [1446864]: segfaults if Tcl_Panic() returns.
| * | | | | | | Decorate Tcl_Panic and Tcl_PanicVA with the noreturn option, alowing further ↵panic_noreturnjan.nijtmans2015-09-227-16/+27
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | optimizations by the C-compiler.
* | | | | | | | micro-opt of hash lookups found by drhMiguel Sofer2015-09-223-13/+11
|/ / / / / / /
* | | | | | | Eliminate compiler warnings for MSVCjan.nijtmans2015-09-221-27/+27
| | | | | | |
* | | | | | | [187d7f499b] Sync the regexp engine to the Postgres version.dgp2015-09-219-147/+182
| | | | | | |
* | | | | | | [1115587][0e0e150e49] Major fix for regexp handling of quantified backrefs.dgp2015-09-214-465/+609
| | | | | | | | | | | | | | | | | | | | | Contributed by Tom Lane from the Postgres project.
* | | | | | | [d7ea9f985][a3c3508599] Internal regexp fixes that cleanup and streamline.dgp2015-09-212-8/+10
|\ \ \ \ \ \ \ | |/ / / / / /
| * | | | | | [d7ea9f985][a3c3508599] Internal regexp fixes that cleanup and streamline.dgp2015-09-212-8/+10
| |\ \ \ \ \ \ | | |_|_|/ / / | |/| | | | |
| | * | | | | [a3c3508599] Fix typo that reversed test sense in prior fix for [3610026].dgp2015-09-211-1/+1
| | | | | | |
| | * | | | | [d7ea9f9853] Contributed patch from tgl with Postgres.dgp2015-09-211-7/+9
| |/ / / / /
| | | | | * Fix for [5d170b5ca5] now available for widespread testing (incl. HPUX and OSX)jan.nijtmans2015-09-071-3/+3
| | | | | |
| | | | | * Merge trunk. jan.nijtmans2015-09-0514-84/+91
| | | | | |\ | |_|_|_|_|/ |/| | | | | "paranoia" patch for the notifier, making sure that thread local data from invalid threads are never accessed by the forked main thread.
* | | | | | Document that multi-threading is on by default since 8.6jan.nijtmans2015-09-041-5/+6
| | | | | |
* | | | | | Remove unnecessary end-of-line spacing (preparing for some further ↵jan.nijtmans2015-09-049-71/+71
| | | | | | | | | | | | | | | | | | | | | | | | documentation updates)
* | | | | | nonportable -> nonPortablejan.nijtmans2015-09-023-8/+8
| | | | | |
| | | | | * In StartNotifierThread() don't lock mutex if thread is already started. jan.nijtmans2015-09-031-50/+22
| | | | | | | | | | | | | | | | | | | | | | | | Fix panic message if thread cannot be started. Remove asserts used for debugging.
| | | | | * Fix the Cygwin notifier, doing the initialization of the thread-local ↵jan.nijtmans2015-09-021-41/+33
| | | | | | | | | | | | | | | | | | | | | | | | variables exactly the same as the Unix notifier.
| | | | | * Merge trunk. jan.nijtmans2015-09-0219-171/+240
| | | | | |\ | |_|_|_|_|/ |/| | | | | Gustaf's latest and greatest fix.
* | | | | | Various Unicode handling enhancements, when building with TCL_UTF_MAX > 3, ↵jan.nijtmans2015-09-013-34/+103
| | | | | | | | | | | | | | | | | | | | | | | | inspired by androwish. No effect if TCL_UTF_MAX=3 (which is the default)
* | | | | | Whitespace reduction in Tcl scripts. No functional change.jan.nijtmans2015-09-0115-135/+135
| | | | | |
| | | | | * Merge trunk. jan.nijtmans2015-08-316-17/+81
| | | | | |\ | |_|_|_|_|/ |/| | | | | Add Gustaf's latest fix for unit-tests.
* | | | | | Some Unicode encoding fixes, only having effect if TCL_UTF_MAX > 4. ↵jan.nijtmans2015-08-312-3/+22
| | | | | | | | | | | | | | | | | | | | | | | | Backported from androwish
* | | | | | [7703ff1082] Improved wording.dkf2015-08-301-4/+5
|\ \ \ \ \ \ | |/ / / / /
| * | | | | [7703ff1082] Improved wording.dkf2015-08-301-4/+5
| | | | | |
* | | | | | Completing [00189c4afc]: Allow semi-static UCRT build on Windows with VC ↵jan.nijtmans2015-08-282-4/+36
|\ \ \ \ \ \ | |/ / / / / | | | | | | | | | | | | 14.0. Now for the configure/makefile build.
| * | | | | Completing [00189c4afc]: Allow semi-static UCRT build on Windows with VC ↵jan.nijtmans2015-08-282-4/+36
| | | | | | | | | | | | | | | | | | | | | | | | 14.0. Now for the configure/makefile build.
* | | | | | Fix [7703ff1082]: extra argument ?pattern? of parray is not documentedjan.nijtmans2015-08-271-8/+9
|\ \ \ \ \ \ | |/ / / / /
| * | | | | Fix [7703ff1082]: extra argument ?pattern? of parray is not documentedjan.nijtmans2015-08-271-8/+9
| | |/ / / | |/| | |
| | | | * merge trunkjan.nijtmans2015-08-254-3/+16
| | | | |\ | |_|_|_|/ |/| | | |
* | | | | Fix [0df7a1ec2674a76a]: unix/tclUnixFile.c: return value of call to ↵jan.nijtmans2015-08-252-3/+6
|\ \ \ \ \ | |/ / / / | | | | | | | | | | TclpGetCwd in TclpFindExecutable is not checked, variable cwd is used uninitialized.
| * | | | Fix [0df7a1ec2674a76a]: unix/tclUnixFile.c: return value of call to ↵jan.nijtmans2015-08-251-2/+5
| | | | | | | | | | | | | | | | | | | | TclpGetCwd in TclpFindExecutable is not checked, variable cwd is used uninitialized.
* | | | | Removed "ucrt" as configureable option: If building with VC2014, it's the ↵jan.nijtmans2015-08-252-22/+4
|\ \ \ \ \ | |/ / / / | | | | | | | | | | recommended way to build Tcl.
| * | | | Removed "ucrt" as configureable option: If building with VC2014, it's the ↵jan.nijtmans2015-08-252-22/+4
| | | | | | | | | | | | | | | | | | | | recommended way to build Tcl.