490372 - improve Mac OS X local file equality testing

Status: RESOLVED FIXED

(Core :: XPCOM, defect)

Description

[Josh Aas]

I think we can improve Mac OS X local file equality testing. Specifically, I don't see why we need to compare FSRef structures which we always need to construct and are never going to use as a path basis in local files. We should just compare CFURLs or native paths.

Comment 1

[Josh Aas]

Attachment: fix v1.0

This just compares native paths. That is almost certainly always going to be a valid path format for comparison, meaning whether we continue to have CFURL or something else (like a native path?) as the basis for our local file store we can always convert to a native path pretty easily.

Comment 2

[Josh Aas]

pushed to mozilla-central

http://hg.mozilla.org/mozilla-central/rev/cb8262ea787f

Comment 3

[Josh Aas]

backed out, unit test orange

http://hg.mozilla.org/mozilla-central/rev/9170bcabf752

I don't understand this error:

TestRegular FAILED - cannot create core service
TestDeferred FAILED - cannot create core service
Finished running RegistrationOrder tests.
make[2]: *** [check] Error 1
make[1]: *** [check] Error 2
make: *** [check] Error 2
program finished with exit code 2

Comment 4

[Josh Aas]

The test is failing because my patch fails to report that the following two paths are equal:

/Users/josh/src/mozilla/ff_192_debug/objdir-debug/dist/bin/components

/Users/josh/src/mozilla/ff_192_debug/objdir-debug/xpcom/tests/../../dist/bin/components

Without my patch these are reported as equal because both directories actually exist and we create an FSRef for each and compare, essentially comparing inodes and not paths.

The files shouldn't have to exist in order for us to compare as far as I can tell, and the UNIX implementation actually does what my first (failed) patch here does and gets away with it. I can only assume that when using the UNIX implementation the paths it is asked to compare aren't the same as the ones I pasted above - they are either both normalized or both non-normalized. I'll test when I have access to Linux again.

Comment 5

[Josh Aas]

Fixed by the patch in bug 491245.