| Commit message (Collapse) | Author | Age | Files | Lines |
| |
|
|
|
|
|
|
| |
Windows caused by missing TCL_THREAD_CREATE_RETURN.
* tests/stack.test (stack-3.1): Corrected nuisance error in
threaded builds.
|
| |
|
|
|
|
| |
for test load-2.3 so that it will pass on Mac OSX.
|
|
|
|
|
| |
compute the target directory, so as not to fail if the
user's HOME isn't the root.
|
| |
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
* generic/tclInterp.c (Tcl_Init): Stopped setting the
tcl_libPath variable. [tclInit] can get all its directories
without it.
* tests/unixInit.test: Modified test code that made use of
tcl_libPath variable.
* unix/tclUnixInit.c: Stopped setting the tclDefaultLibrary variable,
execept on the Mac OS X platform with HAVE_CFBUNDLE. In that configuration
we should seek some way to make use of the TIP 59 facilities and
get rid of that usage of tclDefaultLibrary as well.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
* win/tclWinInit.c (TclpInitLibraryPath): relative-to-executable
directories to the library search path. A first step in reform of
Tcl's startup process.
***POTENTIAL INCOMPATIBILITY***
Attempts to directly run ./tclsh or ./tcltest out of a build
directory will either fail, or will make use of an installed
script library in preference to the one in the source tree.
Use `make shell` or `make runtest` instead.
* tests/unixInit.test: Modified tests to suit above changes.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
build works again.
* generic/tclDecls.h: Changes to the tests for
* generic/tclInt.decls: clock frequency in
* generic/tclIntDecls.h: Tcl_WinTime
* generic/tclIntPlatDecls.h: so that any clock frequency
* generic/tclPlatDecls.h: is accepted provided that
* generic/tclStubInit.c: all CPU's in the system share
* tests/platform.test (platform-1.3): a common chip, and hence,
* win/tclWin32Dll.c (TclWinCPUID): presumably, a common clock.
* win/tclWinTest.c (TestwincpuidCmd) This change necessitated a
* win/tclWinTime.c (Tcl_GetTime): small burst of assembly code
to read CPU ID information, which was added as TclWinCPUID in the
internal Stubs. To test this code in the common case of a
single-processor machine, a 'testwincpuid' command was added to
tclWinTest.c, and a test case in platform.test. Thanks to Jeff
Godfrey and Richard Suchenwirth for reporting this bug. [Bug
#976722]
|
|
|
|
|
| |
we get an exitcode. The windows pipe channel driver doesn't differentiate
between a blocking and non-blocking close just yet, but will soon.
|
| |
|
| |
|
|
|
|
|
|
|
| |
* tests/iogt.test: Tests cases that set invalid -buffersize settings on channels
either wrapped in a catch or modified to a valid range. The assertion results
were not modified to show that the behavior of Tcl_SetChannelBufferSize
is unchanged.
|
|
|
|
| |
if strftime doesn't do %s.
|
|
|
|
|
|
| |
in [clock format], and conditioned out the clock-10.x series if
they're all going to fail because of a broken strftime() call.
[Bug 961714]
|
| |
|
|
|
|
|
|
|
|
| |
* library/tcltest/pkgIndex.tcl: if TCLTEST_OPTIONS value. Corrected
* tests/tcltest.test: double increment of numTestFiles in
-singleproc 1 configurations. Updated tcltest-19.1 to tcltest 2.1
behavior. Corrected tcltest-25.3 to not falsely report a failure
in tcltest.test. Bumped to tcltest 2.2.6. [Bugs 960560, 960926]
|
|
|
|
|
|
| |
* library/tcltest/pkgIndex.tcl: if TCLTEST_OPTIONS value. Updated
* tests/tcltest.test: tcltest-19.1 to tcltest 2.1 behavior.
Bumped to tcltest 2.2.6.
|
|
|
|
|
|
|
|
|
| |
* library/http/http.tcl: that specifies encoding conversion of
* library/http/pkgIndex.tcl: args for http::formatQuery. Previously
* tests/http.test: undefined, RFC 2718 says it should be
utf-8. 'http::config -urlencoding {}' returns previous behavior,
which will throw errors processing non-latin-1 chars.
Bumped http package to 2.5.0.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
* tests/namespace.test (namespace-17.10-12): reverted commit of
2004-05-23 and removed the tests, as it interferes with the
varname resolver and there are apps that break (AlphaTk). A fix
will have to wait for Tcl9.
* generic/tclVar.c: Caching of namespace variables disabled: no
simple way was found to avoid interfering with the resolver's idea
of variable existence. A cached varName may keep a variable's name
in the namespace's hash table, which is the resolver's criterion
for existence.
* tests/namespace.c (namespace-17.10): testing for interference
between varname caching and name resolver.
|
|
|
|
|
|
| |
* tests/unixInit.test: Unset path when done.
* tests/unload.test (unload-3.1): Verify [pkgb_sub] does not exist.
Delete interps when done.
|
|
|
|
|
|
| |
* tests/stringComp.test: stop re-use of string.test test names
* tests/regexpComp.test: stop re-use of regexp.test test names
* tests/namespace.test (namespace-46.3): Verify [p] does not exist.
|
|
|
|
|
|
| |
that it doesn't misdetect some other sort
of filesystem with a write-protected root as
being a CD-ROM drive. [Bug 918267]
|
| |
|
|
|
|
| |
* tests/io.test: Take care to use namespace variables.
|
|
|
|
|
|
|
| |
that gets cleaned up.
* tests/exec.test: Clean up the "path" array.
* tests/interp.test (interp-9.3): Initialize res, so prior values
cannot make the test fail.
|
|
|
|
|
| |
set on ::errorInfo . When left in place, that trace can cause
later tests to fail.
|
| |
|
| |
|
|
|
|
|
|
|
| |
insuring that no "zombie" variables are found.
* generic/tclVar.c (TclLookupSimpleVar): comments re [Bug 736729]
(predecessor of [Bug 959052]) removed.
* tests/namespace.test: added tests 17.10-12
|
|
|
|
|
| |
constraint: [Bug 231259] is closed since nov 2001, and the fix of
[Bug 729692] (INST_START_CMD) makes the test succeed.
|
|
|
|
| |
incredibly constrained.
|
| |
|
| |
|
| |
|
| |
|
| |
|
| |
|
|
|
|
| |
necessary.
|
| |
|
| |
|
| |
|
| |
|
| |
|
| |
|
| |
|
| |
|
| |
|
| |
|