| Commit message (Collapse) | Author | Age | Files | Lines |
|
|
|
| |
port support. [Patch #438509] (schroedter)
|
| |
|
|
|
|
|
| |
option "-error". Essentially ignores the option, always
returning an empty string.
|
|
|
|
|
|
|
|
|
|
| |
* doc/LinkVar.3:
* doc/ObjectType.3:
* doc/PkgRequire.3:
* doc/Preserve.3:
* doc/TCL_MEM_DEBUG.3: Updated documentation to describe the ckalloc,
ckfree, ckrealloc, attemptckalloc, and attemptckrealloc macros, and
to accurately describe when and how they are used. [Bug #497459] (dgp)
|
| |
|
|
|
|
|
|
|
|
| |
Before invoking ckalloc when creating a Tcl_HashEntry,
check that the amount of memory being allocated is
at least as large as sizeof(Tcl_HashEntry). The previous
code was allocating memory regions that were one
or two bytes short. [Bug #521950] (dejong)
|
| |
|
| |
|
| |
|
|
|
|
|
| |
* compat/strtol.c (strtol): gather result in a long before returning
as a long: necessary on platforms where sizeof(int) != sizeof(long).
|
| |
|
|
|
|
| |
have more libdl-like semantics. (bug #514392)
|
| |
|
| |
|
| |
|
|
|
|
|
|
|
| |
* compat/strtoll.c (strtoll):
* compat/strtoul.c (strtoul): Fixed failure to handle leading
sign symbols '+' and '-' and '0X' and raise overflow errors.
[Bug 440916] Also corrects prototype and errno problems.
|
|
|
|
|
|
| |
* unix/tcl.m4 (SC_CONFIG_CFLAGS): Link with -n32
instead of -32 when building on IRIX64-6.* system.
[Tcl bug 521707]
|
|
|
|
|
|
|
| |
* generic/tclObj.c: renamed global variable emptyString ->
tclEmptyString because it is no longer static.
* generic/tclPkg.c: Fix for panic when library is loaded on a
platform without backlinking without proper use of stubs. [Bug 476537]
|
| |
|
|
|
|
| |
in regexp.test for new regsub syntax
|
| |
|
| |
|
|
|
|
|
| |
* unix/tcl.m4: added --enable-64bit support for AIX-4 (using -q64
flag) when using IBM's xlc compiler.
|
| |
|
|
|
|
|
|
| |
sourceable file names (was only *.tcl or tclIndex files with no
more than one dot and 14 chars). There is enough internal
protection in a safe interpreter already. Fixes [Tk Bug #521560].
|
| |
|
| |
|
|
|
|
| |
of address space is quite a lot to play with.
|
| |
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
* win/coffbase.txt (new):
* win/makefile.bc:
* win/makefile.vc: Changed the 'setup' target to stop using
the mkd.bat file and just make the directory right in the rule.
Same change to makefile.bc. configure.in nor Makefile.in use
it.
coffbase.txt will be the master list for our "prefered base
addresses" set by the linker. This should improve load-time
(NT only) by avoiding relocations. Submissions to the list
by extension authors are encouraged.
Added a 'tidy' target to compliment 'clean' and 'hose' to remove
just the outputs. Also removed the $(winlibs) macro as it wasn't
being used.
Stuff left to do:
1) get the winhelp target to stop building in the tools/
directory.
2) stop using rmd.bat
3) add more dependacy rules.
|
|
|
|
| |
easy embedded support. Thanks to Don Porter for pointing this out to me.
|
| |
|
|
|
|
|
|
| |
* win/rules.vc: Added a new "loimpact" option that sets the
-ws:aggressive linker option. Off by default. It's said to
keep the heap use low at the expense of alloc speed.
|
|
|
|
|
|
|
| |
remove the raw windows.h include. tclPort.h brings in windows.h
already and lessens the pre-compiled-header mush and the randomly
useless #pragma comment (lib,...) references throughout the big
windows.h tree (as observed at high linker warning levels).
|
| |
|
| |
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
* win/makefile.vc:
* win/rules.vc: General clean-ups. Added compiler and linker tests
for a) the pentium 0x0F errata, b) optimizing (not all have this),
and c) linker v6 section alignment confusion. All these are tested
first to make sure any D4002 or LNK1117 warnings aren't displayed.
The pentium 0x0F errata is a recommended switch. The v5 linker's
section alignment default is 512, but the v6 linker was changed
to 4096 in an attempt to speed loading on Win98. I changed the
default to always be 512 across both linkers, unless linking
statically, then 4096 is used for the claimed speed effect. Using
a 512 alignment saves 12k bytes of dead space in the DLL.
Added IA64 B-stepping errata switch when the compiler supports it.
Added profiling to $(lflags) when requested and also removed the
explict -entry option as the default works fine as is.
Removed win/tclWinInit.c from the special case section to let it
use the common implicit rule as the $(EXTFLAGS) macro it had was
never referenced anywhere.
|
|
|
|
|
|
|
| |
Overlooked a few source incompatibilities. Now using CONST84.
* generic/tclDecls.h: make genstubs
* generic/tcl.h (Tcl_CmdObjTraceProc): silence warning from Sun
Workshop compiler.
|
| |
|
|
|
|
|
|
|
| |
which should have already been there according to the MSDN docs
regarding the infamous FDIV bug from `97. Added profiling to $(lflags) and
also removed the explict -entry option as the default works without
special work.
|
|
|
|
| |
configure doesn't tell it explicitly.
|
|
|
|
|
|
| |
machine words by default. This seems to be what the test suite expected
anyway, and it is a fairly sensible choice. The other alternative was to
make %d always 32-bit and %ld always 64-bit, but that'd be more complex to do.
|
| |
|
|
|
|
| |
platforms that do not define them.
|
|
|
|
| |
use of VOID* in default case (GNU-ism).
|
| |
|
|
|
|
|
|
|
| |
* tclIOGT.c:
* tclIndexObj.c: Touchups to the TIP 72 patch to make it
compileable under Windows again. The changes are not complete,
there is one nasty regarding _stati64
|
|
|
|
|
| |
This version builds clean on Solaris/SPARC, with GCC and CC, both with and
without threads and both in 32-bit and 64-bit mode.
|