xgettext now required to run tests

[Karen Tracey]

Attempting to run tests on a Windows machine with no xgettext package install now dies a horrible death. The failure to find an xgettext executable causes the whole test run to just quit:

Installed 62 object(s) from 5 fixture(s)
......................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................Error: errors happened while running xgettext on __init__.py
'xgettext' is not recognized as an internal or external command,
operable program or batch file.


C:\u\kmt\django\trunk\tests>

I think we need to somehow figure out if xgettext is available and if not, skip running tests that require it.

[Jannis Leidel]

Hm, does that only occur for Django's tests or for app tests as well?

[Ramiro Morales]

Replying to jezdez:

Hm, does that only occur for Django's tests or for app tests as well?

I suspect it is just the Django test suite with our new makemessages regression test. I can reproduce this by just running it (it seems there are other test failures when xgettext.exe does exist too).

And I suspect it could be related to this:

• No gettext? If you don't have the gettext utilities installed, django-admin.py makemessages will create empty files. If that's the case, either install the gettext utilities or just copy the English message file (locale/en/LC_MESSAGES/django.po) if available and use it as a starting point; it's just an empty translation file. (from ​http://docs.djangoproject.com/en/dev/topics/i18n/localization/#topics-i18n-localization)

• Note: All [os.popen*] functions in this section fail (more or less) silently if the executed program cannot be found; this [subprocess] module raises an OSError exception. (from ​http://docs.python.org/library/subprocess.html#replacing-older-functions-with-the-subprocess-module).

Summary: The old os.popen3 simply ignored the condition of a missing xgettext (or msgmerge or msgfmt) and we used that to our advantage (creating an empty .po file). Now subprocess.Popen() raises an OSError exception instead.

I had seen this and made a mental note about testing this under win32, but forgot to do it later.

Problem was I couldn't get Popen() to raise that OSError under Linux when trying to execute a non-existing external command (e.g. Popen('foo', shell=True, stdout=PIPE, stderr=PIPE, close_fds=False, universal_newlines=True).communicate())

[Jannis Leidel]

Replying to ramiro:

Replying to jezdez:

Hm, does that only occur for Django's tests or for app tests as well?

I suspect it is just the Django test suite with our new makemessages regression test. I can reproduce this by just running it (it seems there are other test failures when xgettext.exe does exist too).

Russ and me we thinking about making the makemessages tests optional just like we do with YAML tests for example, e.g. by a conditional import in the makemessages/tests.py if a call to xgettext succeeds.

And I suspect it could be related to this:

• No gettext? If you don't have the gettext utilities installed, django-admin.py makemessages will create empty files. If that's the case, either install the gettext utilities or just copy the English message file (locale/en/LC_MESSAGES/django.po) if available and use it as a starting point; it's just an empty translation file. (from ​http://docs.djangoproject.com/en/dev/topics/i18n/localization/#topics-i18n-localization)

• Note: All [os.popen*] functions in this section fail (more or less) silently if the executed program cannot be found; this [subprocess] module raises an OSError exception. (from ​http://docs.python.org/library/subprocess.html#replacing-older-functions-with-the-subprocess-module).

Summary: The old os.popen3 simply ignored the condition of a missing xgettext (or msgmerge or msgfmt) and we used that to our advantage (creating an empty .po file). Now subprocess.Popen() raises an OSError exception instead.

I had seen this and made a mental note about testing this under win32, but forgot to do it later.

Problem was I couldn't get Popen() to raise that OSError under Linux when trying to execute a non-existing external command (e.g. Popen('foo', shell=True, stdout=PIPE, stderr=PIPE, close_fds=False, universal_newlines=True).communicate())

Oddly enough Popen() will only raise the OSError for me if I pass it shell=False.

[Jannis Leidel]

Patch to only perform extraction tests if xgettext is found in PATH.

[Jannis Leidel]

(In [12475]) Fixed #12910 - Only test extracting translation strings if xgettext can be found on PATH.

[Jannis Leidel]

(In [12589]) [1.1.X] Fixed #12910 - Only test extracting translation strings if xgettext can be found on PATH.

Backport from r12475.

[Jacob]

Milestone 1.2 deleted