| Commit message (Collapse) | Author | Age | Files | Lines |
|
|
|
| |
::tcltest" to "namespace import -force ::tcltest" [Bug: 3948].
|
| |
|
|
|
|
| |
constraint types to skip certain tests.
|
|
|
|
|
|
| |
constraint types to skip certain tests.
* generic/tclBasic.c (Tcl_SetCommandInfo): comment fix
|
| |
|
| |
|
| |
|
| |
|
|
|
|
|
|
| |
inside of func as it is declared for non-threads builds as well.
In the non-threads case, it always returns TCL_ERROR (couldn't
create thread).
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
* Overall change: Definition of a public API for the creation of
new threads.
* generic/tclInt.h (line 1802f): Removed the definition of
'TclpThreadCreate'. (line 793f) Removed the definition of
'Tcl_ThreadCreateProc'.
* generic/tcl.h (line 388f): Readded the definition of
'Tcl_ThreadCreateProc'. Added Win32 stuff send in by David
Graveraux <davygrvy@bigfoot.com> to that too (__stdcall,
...). Added macros for the default stacksize and allowed flags.
* generic/tcl.decls (line 1356f): Added definition of
'Tcl_CreateThread', slot 393 of the stub table. Two new
arguments in the public API, for stacksize and flags.
* win/tclWinThrd.c:
* mac/tclMacThrd.c: Renamed TclpThreadCreate to Tcl_CreateThread,
added handling of the stacksize. Flags are currently ignored.
* unix/tclUnixThrd.c: See above, but handles joinable
flag. Ignores the specified stacksize if the macro
HAVE_PTHREAD_ATTR_SETSTACKSIZE is not defined.
* generic/tclThreadTest.c (line 363): See below.
* unix/tclUnixNotfy.c (line 210): Adapted to the changes
above. Uses default stacksize and no flags now.
* unic/tcl.m4 (line 382f): Added a check for
'pthread_attr_setstacksize' to detect platforms not implementing
this feature of pthreads. If it is implemented, configure will
define the macro HAVE_PTHREAD_ATTR_SETSTACKSIZE (See
unix/tclUnixThrd.c too).
* doc/Thread.3: Added Tcl_CreateThread and its arguments to the
list of described functions. Removed stuff about not providing a
public C-API for thread-creation.
|
|
|
|
|
| |
dangerous on poorly configured systems [Bug: 3881]
and added 8.2 to keep essence of 8.1 tested.
|
|
|
|
| |
references (no longer supported)
|
|
|
|
| |
[Bug: 3466]
|
|
|
|
|
|
|
|
|
|
|
| |
[Bug: 3466]
* library/tcltest1.0/tcltest.tcl (initConstraints): removed win32s
references (no longer supported)
* tests/fCmd.test: marked test 8.1 knownBug because it is
dangerous on poorly configured systems [Bug: 3881]
and added 8.2 to keep essence of 8.1 tested.
|
| |
|
|
|
|
|
|
|
|
| |
* generic/tclIO.c (Tcl_UnstackChannel, line 1831): Forcing interest
mask to the correct value after an unstack and re-initialization
of the notifier via the watchProc. Without this the first
fileevent after an unstack will come through and be processed,
but no more. [Bug: ??].
|
| |
|
| |
|
| |
|
|
|
|
|
|
|
|
|
|
| |
* library/http2.1/http.tcl: For the -querychannel option,
fconfigure the socket to be binary so that we don't translate
anything while reading the data. This is because we determine the
content length of the data on the channel by using seek (to the end
of the file) and tell on the file handle, and we need the
content-length to match the amount of data actually sent, and
translation can affect the number of bytes posted.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
* Overall change: Definition of public API's for the finalization
of conditions and mutexes. [Bug: 4199].
* generic/tclInt.h: Removed definitions of TclFinalizeMutex and
TclFinalizeCondition.
* generic/tcl.decls: Added declarations of Tcl_MutexFinalize and
Tcl_ConditionFinalize.
* generic/tclThread.c: Renamed TclFinalizeMutex to
Tcl_MutexFinalize. Renamed TclFinalizeCondition to
Tcl_ConditionFinalize.
* generic/tclNotify.c: Changed usage of TclFinalizeMutex to
Tcl_MutexFinalize.
* unix/tclUnixNotfy.c:
* generic/tclThreadTest.c: Changed usages of TclFinalizeCondition to
Tcl_ConditionFinalize.
* generic/tcl.h: Added empty macros for Tcl_MutexFinalize and
Tcl_ConditionFinalize, to be used when the core is compiled
without threads.
* doc/Thread.3: Added description the new API's.
|
| |
|
|
|
|
|
|
|
| |
correctly [Bug: 4085]
* unix/tclUnixFCmd.c (SetGroupAttribute):
* unix/tclUnixFCmd.c (SetOwnerAttribute): Added (uid_t) and (gid_t)
casts to avoid compiler warnings.
|
|
|
|
| |
improved translation to winhelp [Bug: 3679]
|
|
|
|
|
|
| |
* tests/namespace.test:
* tests/var.test: Added lsorts to avoid random sorted return
problems. [Bug: 2682]
|
|
|
|
| |
to prevent itcl info override crash [Bug: 4064]
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
to prevent itcl info override crash [Bug: 4064]
* tests/foreach.test:
* tests/namespace.test:
* tests/var.test: Added lsorts to avoid random sorted return
problems. [Bug: 2682]
* tests/fileName.test: fixed 14.1 test fragility [Bug: 1482]
* tools/man2help2.tcl: fixed winhelp cross-linking error [Bug: 4156]
improved translation to winhelp [Bug: 3679]
* unix/Makefile.in (MAN_INSTALL_DIR): patch to accept --mandir
correctly [Bug: 4085]
* unix/dltest/pkg[a-e].c: Cleaned up test packages [Bug: 2293]
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
error test (previously only errno was checked, but the return
value of strtod() should be checked as well). [Bug: 4118].
* tests/exec.test: Added test for proper conversion of UTF data
when used with "<< $dataWithUTF" on exec's.
* unix/tclUnixPipe.c (TclpCreateTempFile): Added
Tcl_UtfToExternalDString call, so that if there is UTF content in
the string it will be properly converted to the system encoding
before being written [Bug: 4030].
(TclpCreateTempFile): Added a check on the return value of tmpnam;
some systems (Linux, for example) will start to return NULL after
tmpnam has been called TMP_MAX times; not checking for this can
have bad results (overwriting temp files, core dumps, etc.)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
* generic/tclIntPlatDecls.h:
* generic/tclStubInit.c:
* win/tclWin32Dll.c: removed TclWinSynchSpawn (vestige of Win32s
support).
* win/tclWinReg.c: made use of TclWinGetPlatformId instead of
getting info again
* win/tclWinPort.h:
* win/Makefile.in:
* win/configure.in:
* win/tcl.m4: Added support for gcc/mingw on Windows [Bug: 4234]
|
|
|
|
|
|
|
|
|
|
| |
noting the need to pair ckalloc with ckfree. [Bug: 4262]
* generic/tclInt.decls:
* generic/tclIntPlatDecls.h:
* generic/tclStubInit.c:
* win/tclWin32Dll.c: removed TclWinSynchSpawn (vestige of Win32s
support).
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
noting the need to pair ckalloc with ckfree. [Bug: 4262]
* generic/tclInt.decls:
* generic/tclIntPlatDecls.h:
* generic/tclStubInit.c:
* win/tclWin32Dll.c: removed TclWinSynchSpawn (vestige of Win32s
support).
* win/tclWinReg.c: made use of TclWinGetPlatformId instead of
getting info again
* win/tclWinPort.h:
* win/Makefile.in:
* win/configure.in:
* win/tcl.m4: Added support for gcc/mingw on Windows [Bug: 4234]
|
|
|
|
|
|
|
|
| |
more aware of TCL_BYTECODE_PRECOMPILED flagged structs (gen'd by
tbcload), to correctly clean them up.
* generic/tclClock.c (FormatClock): moved check for empty format
earlier, commented 0 result return value
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
* library/http2.1/http.tcl: Removed an unnecessary fileevent
statement from the error processing part of the Write method.
Also, fixed two potential memory leaks in wait and reset, in which
the state array wasn't being unset before throwing an exception.
Prior to this version, Brent checked in a fix to catch a
fileevent statement that was sometimes causing a stack trace when
geturl was called with -timeout. I believe Brent's fix is
necessary because TLS closes bad sockets for secure connections,
and the fileevent was trying to act on a socket that no longer
existed.
|
|
|
|
|
|
|
| |
* tests/namespace.test:
* generic/tclNamesp.c (Tcl_Export): added a uniq'ing test to the
export list so only one instance of each export pattern would
exist in the list.
|
|
|
|
|
|
|
|
|
| |
* generic/tclNamesp.c (Tcl_Export): added a uniq'ing test to the
export list so only one instance of each export pattern would
exist in the list.
* generic/tclExecute.c (TclExecuteByteCode): optimized case for
the empty string in ==/!= comparisons
|
|
|
|
|
|
|
|
|
|
| |
* tests/namespace.test:
* generic/tclNamesp.c (Tcl_Export): added a uniq'ing test to the
export list so only one instance of each export pattern would
exist in the list.
* generic/tclExecute.c (TclExecuteByteCode): optimized case for
the empty string in ==/!= comparisons
|
| |
|
|
|
|
|
|
|
|
| |
[Bug: 4409].
* unix/tclLoadAout.c:
* unix/tclUnixPipe.c: Added (off_t) type casts in lseek() calls
[Bug: 4410].
|
| |
|
|
|
|
|
|
| |
* library/http2.1/http.tcl: Fixed a bug where string query data
that was bigger than queryblocksize would get duplicate characters
at block boundaries.
|
|
|
|
|
|
|
|
|
|
| |
* library/http2.1/http.tcl: Fixed bug 4463, where we were getting
a stack trace if we tried to publish a project to a good host but
a port where there was no server listening. It turned out the
problem was a stray fileevent that needed to be cleared. Also,
fixed a bug where http::code could stack trace if called on a bad
token (one which didn't represent a successful geturl) by adding
an http element to the state array in geturl.
|
|
|
|
|
|
|
|
|
| |
respect to the time zone in which they were run and were thus
failing.
* doc/clock.n: Clarified meaning of -gmt with respect to -base
when used with [clock scan] (-gmt does not affect the
interpretation of -base).
|
|
|
|
|
|
|
|
| |
* library/http2.1/http.tcl: geturl used to throw an exception when
the connection failed; I accidentally returned a token with the
error info, breaking backwards compatibility. I changed it back
to throwing an exception, but unsetting the state array first
(thus still eliminating the original memory leak problem).
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
some of Brent's modifications to allow asynchronous posts (via
-command). Also modified -queryprogress so that it calls the
query callback as <callback> <token> <total size> <current size>
to be consistent with -progress. Added -queryblocksize option
with default 8192 bytes for post blocksize. Fixed a bunch of
potential memory leaks for the case when geturl receives bad args
or can't open a socket, etc. Overall, the package really rocks
now.
* doc/http.n: Added -queryblocksize, -querychannel, and
-queryprogress. Also, changed the description of -blocksize,
which states that the -progress callback will be called for each
block, to now qualify that with an "if -progress is specified".
* tests/http.test: Added a querychannel test for synchronous and
asynchronous posts, altered the queryprogress test such that the
callback conforms to the -progress format. Also, had to use the
-queryblocksize option to do the post 16K at a time to match
Brent's expected results (and to test that -queryblocksize works).
|
| |
|
| |
|
|
|
|
|
|
|
|
| |
http::geturl and also changed it so that writing the post data
is event driven if the queryprogress callback or a timeout is given.
This allows a timeout to occur when writing lots of post data.
The queryprogress callback is called after each block of query
data is posted. It has the same signature as the -progress callback.
|
| |
|
|
|
|
| |
.dlls
|
|
|
|
|
|
|
| |
setting geometry of slave interp to 0x0 when Tk was loaded, it now
does "wm withdraw .". Both remove the main window from the
display, but the former caused some internal structures to get
initialized to zero, which caused crashes with some extensions.
|