| Commit message (Collapse) | Author | Age | Files | Lines |
... | |
| | | | | |
|
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
This change adds support for scanning multiple entries in an action string in
order to better support the following use cases:
1. A file is provided in an action string and should be taken as a dependency.
For example, an action string "$PERL somefile.pl".
2. An action string actually has two actions separated by &&. For example,
"cd <some_dir> && $ZIP <args>".
Adding support for #1 actually allows us to fix the test
IMPLICIT_COMMAND_DEPENDENCIES.py on Windows, which was previously treating a
Python file as executable even on Windows. This was causing tests to repeatedly
open the default handler of Python files, which if set to Visual Studio causes
DDE hangs. This test is fixed because now we can have the action string specify
python as the first command and still take an implicit dependency on the
script, which is now the second command.
|
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
In one location a copy-paste apparently didn't get fixed up,
and the wrong construction var was used.
Fixes #3540
Signed-off-by: Mats Wichmann <mats@linux.com>
|
|\ \ \ \ \
| | |_|_|/
| |/| | | |
|
| |\ \ \ \
| | | | | |
| | | | | | |
More post py27 cleanup
|
| | | | | | |
|
| | | | | | |
|
| | | | | |
| | | | | |
| | | | | |
| | | | | | |
in test to complete
|
| |\ \ \ \ \
| | |/ / / / |
|
| | | | | | |
|
| | | | | | |
|
| | | | | | |
|
| | | | | | |
|
| | | | | | |
|
| | | | | | |
|
| | | | | | |
|
| | | | | | |
|
| | | | | | |
|
| | | | | | |
|
| | | | | | |
|
| | | | | | |
|
| | | | | | |
|
| | | | | |
| | | | | |
| | | | | |
| | | | | | |
dropping < py 3.5
|
| | | | | | |
|
| | | | | | |
|
| | |/ / /
| | | | |
| | | | |
| | | | | |
contexts are opened
|
| |\ \ \ \
| | |/ / /
| |/| | | |
|
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
Also simplify cleanup a bit - we don't need to leave the site file open,
can use a context manager to read the code and let it close there.
Signed-off-by: Mats Wichmann <mats@linux.com>
|
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
instead of deprecated imp module.
import site_init.py file directly for Py3
Find spec for the site file by path rather than by fiddling
sys.path to restrict the search.
Py3: address the problem that we might get a "site_init.py" from
somewhere else in sys.path.
Clean up excess indentation / try nesting; adjust comment.
Signed-off-by: Mats Wichmann <mats@linux.com>
|
| | | | | |
|
| | | | | |
|
|\ \ \ \ \
| |/ / / / |
|
| |\ \ \ \
| | | | | |
| | | | | | |
Add a warning if msvc config cache may be outdated.
|
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | | |
Be more clear about the msvc config-cache issue being fixed.
Signed-off-by: Mats Wichmann <mats@linux.com>
|
| | |/ / /
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
If we didn't find cl.exe in the final check in msvc_setup_env(),
issue a different warning if config caching is enabled. If it is,
there's a decent chance we've found cl.exe in the past, meaning
the cache is probably out of date - entries are indexed by path
to the bat file+arg, whereas an msvc update may bump a version
number in the path and the old path won't be valid: a cached
path entry could look like this:
"C:\\Program Files (x86)\\Microsoft Visual Studio\\2019\\Community\\VC\\Tools\\MSVC\\14.24.28314\\bin\\HostX64\\x64",
and the .28314 could be changed in an update.
Signed-off-by: Mats Wichmann <mats@linux.com>
|
| | | | | |
|
| | | | |
| | | | |
| | | | |
| | | | | |
are interpreted instead of simply used to replace text. Switch to using str.replace()
|
| |/ / / |
|
| |\ \ \
| | | | |
| | | | | |
Add support for VS2017 Express
|
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
Chocolatey install path is now checked. The paths to try use a
different method of getting the Windows vars - os.path.expandvars,
that avoids the use of a try block or os.environ.get.
Signed-off-by: Mats Wichmann <mats@linux.com>
|
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
During debugging, the call to vswhere which uses a derived
version from the table was replaced by a fixed range, and
this was unfortunately committed to git. Restoring.
Signed-off-by: Mats Wichmann <mats@linux.com>
|
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
subprocess.run kw arg capture_output actually doesn't appear
until Python 3.7; drop back to individually setting stdout/stderr.
Signed-off-by: Mats Wichmann <mats@linux.com>
|
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
Make tests still pass when run from a non-Windows platform, was
breaking in CI on Linux hosts.
Fiddle some more with comments.
Add 14.1Exp in more places, including docs.
Signed-off-by: Mats Wichmann <mats@linux.com>
|
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
For Express 2017, no 64-bit tools. When that was fixed to
retry with 32-bit tools, the case where a 64-bit target
was specified failed.
Signed-off-by: Mats Wichmann <mats@linux.com>
|
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
Some changes in vswhere detection - not specific to 2017 Express,
but came up in testing so fixing at same time: use the Windows
variables for where programs are installed, rather than hardcoding C:.
[fixes #3542]
Since py2 support is dropped, use the more modern subprocess.run
to call vswhere instead of fidlling with a Popen object (change
is minimal, but this eliminates the commented aside about Py2 not
supporting a context manager on Popen instances, as run() is a complete
implementation of setup-call-wait-teardown).
Signed-off-by: Mats Wichmann <mats@linux.com>
|
| | | | |
| | | | |
| | | | |
| | | | | |
Signed-off-by: Mats Wichmann <mats@linux.com>
|
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
VS 14.1+ looks for a host/target batch file, except those
didn't exist in that form if host=target. add the
ability to find those (vcvars32.bat and vcvars64.bat)
Signed-off-by: Mats Wichmann <mats@linux.com>
|
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
Needed some new logic to accomodate that tools are not in the expected
place (if real host is amd64, Express still uses Hostx86 as the tools
dir, not Hostx64).
Also needed a hack to run vcvarsall.bat with the correct argument. We
now derive the "argument" by finding the appropriate host_target
vcvars script, which has the argument to vcvarsall embedded, which
we then run instead of calling vcvarsall directly. This allows us to
deal with the pseudo target that Express has always used - we derive
this part of the script name by using an existing lookup table.
arm targets were also added (untested).
Added some pithy comments (oh, no, the Piths ran away!)
Signed-off-by: Mats Wichmann <mats@linux.com>
|
| | | | |
| | | | |
| | | | |
| | | | | |
list of developers
|
| | | | |
| | | | |
| | | | |
| | | | | |
change info to file
|