[modules-dev] Review request for 6608500
Dave Bristor
David.Bristor at Sun.COM
Wed Sep 26 16:25:14 PDT 2007
Stanley M. Ho wrote:
>
> Dave Bristor wrote:
>> This is the prior in behavior, unchanged. I changed code so it's now
>> in temp. But then I observed that, since we don't set the property in
>> all our tests, an expansion directory is created for each test in
>> temp! Kelly / JPRT won't be pleased, and that's contrary to the point
>> of this bugfix. I propose to revert the code for now, and cause the
>> expansion directory to be created on an as-needed basis, during
>> install(), as we discussed yesterday w.r.t. the other bug on my plate,
>> 6605100. An alternative would be to modify all of our tests to set
>> the extension directory's path as part of this fix. The webrev has
>> the code reverted to using source directory.
>
> I'm okay if we revert the code for now, but I think we should definitely
> address it in a bug fix soon. Regardless, expanding into the source
> directory won't be the correct behavior; the expanded directory should
> be either somewhere in temp, or in the repository's cache defined in the
> implementation.
>
> I think the real problem in the tests is that we don't remove the
> expansion directory after each test is executed. Modifying all of our
> tests to set the expansion directory would be one way. An alternative is
> to call Repository.shutdown() before each test exits (or implement the
> shutdownOnExit() method), and the repository's shutdown could clean up
> the expansion directory automatically.
I think we have to wait for releaseModule / disableModuleDefinition before
this will work: tests on Windows fail because some code has a file descriptor
open. URLRepository.shutdown() does try to remove them but does not ascertain
that they have been removed. LocalRepository.shutdown() doesn't have file
removal yet implemented, but we have 6559067 which I can do Real Soon Now.
Thanks,
Dave
>
> - Stanley
More information about the modules-dev
mailing list