551658 - [@ nsIFile::createUnique] crashes with long parent directory names

Status: RESOLVED FIXED

(Core :: XPCOM, defect)

Description

[:Paolo Amadini]

Attachment: Patch and test case, tested on Windows only

nsIFile::createUnique may crash with long parent directory names.
See the attached patch and test case for details.

Unfortunately I can't compile nor test the part of the patch for operating
systems other than Windows (after the #else), I'd need help from someone
who can build on other platforms.

Comment 1

[:Paolo Amadini]

Thanks for the review! Just to be safe before check-in, did you verify
that this compiles on at least one platform other than Windows?

Comment 2

[timeless]

oh, i should note that i think we should probably not go this way :).

Mozilla no longer supports w9x, which means we should be able to use \\?\ notation if we need long file paths....

Comment 3

[:Paolo Amadini]

(In reply to comment #1)
> Thanks for the review! Just to be safe before check-in, did you verify
> that this compiles on at least one platform other than Windows?

I guess I'll just mark this for check-in with a note.

Comment 4

[Kyle Huey (Exited; not receiving bugmail, old account, do not use)]

I threw this at tryserver, will post the results.

Comment 5

[timeless]

again, this is really the wrong way to fix the problem.

Comment 6

[:Paolo Amadini]

(In reply to comment #2)
> Mozilla no longer supports w9x, which means we should be able to use \\?\
> notation if we need long file paths....

Although the Windows API allows for the creation of files with very long
names, doing this is not recommended, as the both the shell and most command
line tools will generally refuse to work with paths longer than 260 characters.

createUnique is used by the file download code to generate a unique name when
downloading automatically, and we don't want such downloaded file to become
unmanageable if a long name is suggested by a "Content-Disposition" header.

Besides, I don't know how well long names are handled on platforms other
than Windows.

Comment 7

[timeless]

"fairly well". minus some stupid issues that are really outside of scope (posix sucks)

Comment 8

[:Paolo Amadini]

(In reply to comment #4)
> I threw this at tryserver, will post the results.

Seems like the tryserver build succeeded?

http://tinderbox.mozilla.org/showbuilds.cgi?tree=MozillaTry&maxdate=1269813016&hours=24&legend=0&norules=1

Comment 9

[:Paolo Amadini]

Still waiting for check-in?

Comment 10

[Dão Gottwald [:dao]]

http://hg.mozilla.org/mozilla-central/rev/739a856b24eb