summaryrefslogtreecommitdiffstats
path: root/generic
Commit message (Collapse)AuthorAgeFilesLines
* Another round of sentinel fixes for 8.6, NULL -> (char *)NULLjan.nijtmans2024-05-2935-217/+217
|
* Make TclGlob() a static function. Sentinel/indenting/comment improvements ↵jan.nijtmans2024-05-2819-562/+560
| | | | (all backported from 8.7)
* Eliminate TclPrintByteCodeObj()'s 'interp' argument, which is not usedjan.nijtmans2024-05-275-121/+99
|
* fix for [e3f4a8b78dec4bdb]: don't swallow limit errors in further ↵sebres2024-05-272-13/+4
| | | | | invocations of EvalObjvCore (e. g. direct invocation or NRE, Tcl_EvalObjv, Tcl_EvalObjEx, TclNREvalObjEx, etc); partially reverts [b740e2abbd44c7d0]
* More C functions for working with dicts [656fe3c816]dkf2024-05-2615-148/+237
|\
| * Use the new operationsdkf2024-05-2515-148/+117
| |
| * Add some more C functions for working with dicts [656fe3c816]dkf2024-05-252-0/+120
| |
* | Check limits immediately when we do [interp eval]. [e3f4a8b78d]dkf2024-05-251-0/+13
|/
* Use TclDStringToObj a bit moredkf2024-05-241-5/+1
|
* Code optimization: no need for encoding = Tcl_GetEncoding(NULL, NULL). Use ↵jan.nijtmans2024-05-241-2/+1
| | | | TclDStringToObj where possible
* Fix [3fc3287497]: TclGetProcessGlobalValue encodes information twice on Windowsjan.nijtmans2024-05-233-6/+14
|\
| * Possible fix for [3fc3287497]: TclGetProcessGlobalValue encodes information ↵jan.nijtmans2024-05-231-4/+12
| | | | | | | | twice on Windows
* | Fix 2 warnings on Win32 (Thanks, Harald). Some more code cleanup, backported ↵jan.nijtmans2024-05-232-74/+67
|/ | | | from 8.7)
* cherry-pick [659ca0ae8da43a1e] for 8.6: don't need to invoke it in case if ↵sebres2024-05-221-0/+3
| | | | oPtr->selfCls is NULL
* closes [79474c58800cdf94]: fixes 2 segfaults and 2 leaks (common IO handlers ↵sebres2024-05-212-41/+111
|\ | | | | | | and reflected channels)
| * merge apn-channelstate-leak, fixes possible leaks on several members that ↵sebres2024-05-171-1/+29
| |\ | | | | | | | | | may be set after the close, asynchronously or via some callbacks or handlers
| | * Free ChannelState.chanMsg .unreportedMsgapnadkarni2024-05-171-2/+20
| | |
| * | fixes yet another segfault (if write handler doesn't generate an error, but ↵sebres2024-05-171-8/+9
| |/ | | | | | | returns normally with length of written bytes)
| * fix mem-leak originating by cyclic reference `rcPtr->name (type "channel", ↵sebres2024-05-161-19/+22
| | | | | | | | | | its refCount may be larger than 1) => statPtr => chanPtr => chanPtr->instanceData => refChan`: this would avoid that object rcPtr->name (name of channel that gets deleted or dead) still holds the reference to statPtr, see 2nd part of bug [79474c58800cdf94].
| * A couple of tiny code style improvementsdkf2024-05-161-4/+4
| |
| * code review: preserving r/w channels after all possible failures (don't need ↵sebres2024-05-161-3/+3
| |\ | | | | | | | | | to release in error case)
| | * code review: preserving r/w channels after all possible failures (don't need ↵sebres2024-05-161-3/+3
| | | | | | | | | | | | to release in error case)
| * | small code review: removal of unneeded - result is initially TCL_OKsebres2024-05-151-3/+0
| |\ \ | | |/
| | * small code review: removal of unneeded - result is initially TCL_OKsebres2024-05-151-3/+0
| | |
| * | ultimate fix for [79474c58800cdf94]: avoid segfault on copy-state structure ↵sebres2024-05-151-13/+54
| |\ \ | | |/ | | | | | | freed to early, protected by refCount and by preserving its r/w channels now
| | * ultimate fix for [79474c58800cdf94]: avoid segfault on copy-state structure ↵sebres2024-05-151-13/+54
| | | | | | | | | | | | | | | freed to early, protected by refCount and by preserving its r/w channels now; back-ported (squashed rewrite) from tclSE to 8.5 base
* | | Fix for [7842f33a5c]: Stereotype call chains were ending up bogus in some ↵dkf2024-05-201-6/+52
| | | | | | | | | | | | situations
* | | Remove the band-aiddkf2024-05-171-6/+1
| | |
* | | Fix the problem properlydkf2024-05-172-103/+56
| | |
* | | Starting to clean up the mess. The extra frame info can have the same ↵dkf2024-05-172-54/+91
| | | | | | | | | | | | lifespan as the method itself.
* | | Apply an emergency workarounddkf2024-05-161-1/+6
|/ /
* | issue [10aa8403d064439f]: make command `testsize` platform-independent ↵sebres2024-05-101-0/+23
| | | | | | | | (covers constraint time64bit for unix too)
* | Remove COMPAT==0 part, no longer makes sense. More code-cleanup, backported ↵jan.nijtmans2024-05-022-158/+76
| | | | | | | | from 8.7
* | Make sure to forward-declare "struct addrinfo". Some compilers don't like ↵jan.nijtmans2024-05-021-0/+1
| | | | | | | | doing that in a parameter-list.
* | TCL_NORETURN1 is empty for extensions, just not for Tcl and Tk builds. Why? ↵jan.nijtmans2024-04-295-18/+12
| | | | | | | | Backport changing this from 8.7/9.0
* | Missing TCL_NORETURN for Tcl_ExitThread() (and TclpThreadExit()). Backported ↵jan.nijtmans2024-04-297-24/+24
| | | | | | | | from 8.7/9.0.
* | __declspec should have double underscore (even though it works with one)jan.nijtmans2024-04-241-1/+1
| |
* | Sync parameter name in tcl.decls with other placesjan.nijtmans2024-04-212-3/+3
| |
* | targetNamePtr -> targetCmdPtr (making it the same as in other documentation ↵jan.nijtmans2024-04-191-10/+10
| | | | | | | | and tcl.decls)
* | Reserve 2 more stub entries for TIP #648jan.nijtmans2024-04-193-5/+13
| |
* | More code cleanup, backported from 8.7jan.nijtmans2024-04-192-278/+283
| |
* | Reduce the diff between 8.6 and 8.7, by backporting some ↵jan.nijtmans2024-04-182-144/+161
| | | | | | | | formatting/restructuring. No functional change.
* | explaination commentsebres2024-04-121-1/+1
| |
* | merge 8.5 (minor backport from my core) simple speed-up if searching for the ↵sebres2024-04-121-1/+4
|\ \ | |/ | | | | key from hash itself (it is safe to compare needle with an address to key.string)
| * minor backport from my core: simple speed-up if searching for the key from ↵sebres2024-04-121-30/+32
| | | | | | | | | | hash itself (it is safe to compare needle with an address to key.string); back to mixed identation (removed before)
* | merge 8.5, amend to [82f721a5a039596f]: the bug was fixed incompletely, this ↵sebres2024-04-125-15/+39
|\ \ | |/ | | | | | | | | is full bug fix now - don't allow direct compare if keys contain values rather than pointers; introduced new hash-key type flag TCL_HASH_KEY_DIRECT_COMPARE... I know it is public interface, but the bug is grave, and I don't know how one could fix it without that, by retaining same performance for pointer hashes (e. g. vars, dicts and all of TclObjs).
| * amend to [295b0570ff660950]: the bug was fixed incompletely, this is full ↵sebres2024-04-125-18/+45
| | | | | | | | | | | | bug fix now - don't allow direct compare if keys contain values rather than pointers. introduced new hash-key type flag TCL_HASH_KEY_DIRECT_COMPARE... I know it is public interface, but the bug is grave, and I don't know how one could fix it without that, by retaining same performance for pointer hashes (e. g. vars, dicts and all of TclObjs).
* | Re-generate tclTomMath.h from tommath.hjan.nijtmans2024-04-121-31/+41
| |
* | Use Tcl_NewBooleanObj for booleans; we should say what we mean. Fix ↵jan.nijtmans2024-04-1115-198/+202
| | | | | | | | sentinels and type-casts.
* | Tcl_NewBooleanObj() is better here than Tcl_NewIntObjjan.nijtmans2024-04-101-1/+1
| |
generated by cgit v0.12 at 2024-11-16 01:33:11 (GMT)