| Commit message (Collapse) | Author | Age | Files | Lines |
|---|
| ... | |
| |
|
|
|
|
|
|
|
| |
so the user can define a path to look for makeindex style files.
When the latex files are generated by another builder the user
may have to define env['TEXINPUTS'] or some of the other
environment variables so latex can find files that are
not sideeffects. This woul dinclude style files and the like.
|
| |
|
|
| |
most of the time but it needed fixing.
|
| |
|
|
| |
Sorry about the multiple commits!!
|
| |
|
|
| |
file copy and rename so it is same for Windows, Linux and Mac
|
| |
|
|
|
| |
Related fixes: explicitly close the db when we close it,
and add a .close() method to dblite.py.
|
| |
|
|
|
| |
(specifically 'MD5' and 'timestamp-match'), copied from back when
this functionality was configured with the SourceSignatures() function.
|
| |
|
|
|
|
|
| |
add them if the .tex file is generated by another builder and is
not yet created.
If the file exists then only add the side effects if they will be created.
|
| | |
|
| |
|
|
| |
Windows, and always use -D for CPPDEFPREFIX for simplicity.
|
| |
|
|
| |
Jean-Baptiste Lab. Fixes issue #2671.
|
| |
|
|
| |
#1938.
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| |
|
|
| |
From Jean-Franois Colson.
|
| |
|
|
| |
issue 2573.
|
| |
|
|
| |
closes issue 2520.
|
| | |
|
| | |
|
| | |
|
| | |
|
| |
|
|
| |
the children, including the explicit dependencies, not just the sources.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
http://scons.tigris.org/svn/scons/branches/pending
........
r4942 | stevenknight | 2010-06-03 12:41:20 -0700 (Thu, 03 Jun 2010) | 13 lines
Isseu 2641: Latest drop of the TestCmd infrastructure, v. 1.3, including:
* Support for test timeouts.
* Ability to set separate match_stdout and match_stderr functions.
* Ability to set separate diff_stdout and diff_stderr functions.
* Static methods for the various underlying match* and diff* functionality.
* Ability to get at the various match* and diff* functions by attribute name.
* Got rid of checks for difflib now that Python 2.3 is the floor
(for this infrastructure, anyway).
Ripple effects in two test scripts. Added upstream unit test modules
(QMTest/Test{Cmd,Common}Tests.py). Added a README.txt file.
........
r4943 | stevenknight | 2010-06-03 13:00:31 -0700 (Thu, 03 Jun 2010) | 2 lines
Grab the correct TestCmd files with the updated version number.
........
r4946 | managan | 2010-06-04 09:39:20 -0700 (Fri, 04 Jun 2010) | 4 lines
On Windows add a '/D' to the command line so it recognizes drive letters in the
source or target file paths
........
r4947 | managan | 2010-06-04 09:51:48 -0700 (Fri, 04 Jun 2010) | 5 lines
The scanner was not parsing the dependencies in \includegraphics commands when there was whitespace (including carriage returns) in the command.
While we need a better long term fix this covers this concern.
........
r4948 | managan | 2010-06-04 11:13:12 -0700 (Fri, 04 Jun 2010) | 3 lines
Dropped an import line that is needed by the last commit of mine for Windows
depenedant option on latex command lines
........
r4949 | managan | 2010-06-04 12:27:48 -0700 (Fri, 04 Jun 2010) | 7 lines
Some latex packages break up commands where you normally could not by
using a comment character at the end of the first line.
Our current scanner broke on this and lost some dependecies.
While we need general fix, this patch solves this problem
........
r4950 | managan | 2010-06-04 15:51:36 -0700 (Fri, 04 Jun 2010) | 5 lines
Tweak how we handle comments within Latex source files when
scanning and looking for dependencies. We were adding a
space when a comment broke a line and we should not have.
........
r4984 | managan | 2010-06-07 09:37:40 -0700 (Mon, 07 Jun 2010) | 6 lines
The multi-line_include-options test failed to check for the
existence of latex. Added that so this test is skipped on
systems without latex.
........
|
| | |
|
| | |
|
| | |
|
| |
|
|
|
|
|
|
|
|
|
|
| |
(--interactive mode on our Solaris buildbot slave takes longer than
10 seconds to create the necessary marker files.)
Set the stdin attribute on our popen object to None after we close it,
so some versions of subprocess.py don't try to write it after it's closed.
If we time out waiting for a marker file to be created, then print any
stdout or stderr that's already accumulated, since that may give clues
about why the test failed.
|
| | |
|
| |
|
|
|
|
|
|
|
| |
* Different messages and exit codes for errors.
* Fix Solaris shared suffixes.
* Add a function declaration to some in-line C code to avoid a warning
from Sun's compiler.
* Make test/Parallel/multiple-parents.py chatty to avoid hangs, and
reduce how many times it calls SCons just to make it go a little quicker.
|
| | |
|
| |
|
|
| |
intermediate messages to avoid buildbot hangs.
|
| |
|
|
|
|
|
|
|
|
|
| |
platform; it was causing the test to fail on Windows.
Added some documentation to bin/update-release-info.py.
In my last log message, I forgot to mention that QMTest/TestRuntest.py was
tweaked to make it less specialized for testing runtest.py. It can now be
used to test other programs; in this case, bin/update-release-info.py. The
defaults are unchanged so that it will normally test runtest.py.
|
| | |
|
| | |
|
| |
|
|
|
|
|
|
|
|
|
|
| |
This program automates the process of inserting the release configuration
information that is currently done by hand. The configuration data now
all lives in a file, 'ReleaseConfig'. Change the configuration file, run
the program, and, hey presto, the files are updated.
More work needs to be done; the documentation doesn't quite match with the
program and the regression tests only cover the release flows, but it's
important to get this stuff out before the next release is published so
people can start using it.
|
| |
|
|
|
| |
'msvc', 'mslink', 'midl'], so the test can actually find and compile
against things like the atlbase.h file.
|
| |
|
|
| |
of non-zero exit status on systems that don't have os.WEXITSTATUS().
|
| | |
|
| |
|
|
| |
This should alter the HOST_ARCH for these tests on non-x86 platforms to x86 and therefore run.
|
| | |
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
| |
to go.
Update various files to have the correct date, time, version, and floor.
Rework Script/Main.py and QMTest/TestSCons.py to make automated updating
possible in the future.
Remove a duplicate MANIFEST-xml.in entry.
Fix TeX tests, which ran 'kpsewitch' before determining if the TeX tools
were installed.
|
| |
|
|
|
| |
Thanks to Steven Knight, the SCCS and Perforce tests have been debugged. Only
the BitKeeper test remains to be debugged.
|
| |
|
|
|
|
| |
Start the deprecation cycle for SourceCode() and its related factory functions.
Deprecation messages are added to the functions. Regression tests are modified
to deal with the messages being generated. Documentation is updated.
|
| |
|
|
|
|
|
| |
* incorporate and update the Python 3.0 fixer work by Greg:
* full conversion to using subprocess for execution;
* rename exactly_contain_all_lines() to must_contain_exactly_lines();
* other minor fixes and stylistic cleanups.
|
| |
|
|
| |
the deprecation cycle.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
keyword parameter.
Several existing tests were still using BuildDir() or build_dir; they were
converted to use VariantDir() and variant_dir.
New tests were added to validate that the --warn=deprecated-build-dir option
and the SetOption method did the right thing. This led to the discovery that
a commonly-used test pattern provided by the infrastructure gobbled up too
much, causing tests to succeed when they should have failed. Fixing the
pattern led to other tests needing to be fixed.
In the process, it was discovered that the SCONSFLAG environment variable was
not getting correctly reset to its original value. Fixing this also caused
additional tests to misbehave, requiring them to be updated.
And test/Sig.py, which tests the deprecated SCons.Sig module, was moved to
the test/Deprecated directory.
All in all, quite a lot of action for what was supposed to be a simple change.
|
| | |
|
| | |
|
