Bug 219572 – [WebCore] Fix build without PATH_MAX

• Status: NEW
• Component: WebCore Misc.

[Samuel Thibault]

Created attachment 415505 [details]
proposed fix

This fixes build on GNU/Hurd, which does not have a PATH_MAX limitation.

[Samuel Thibault]

Created attachment 415526 [details]
Patch

[Don Olmstead]

I feel like it'd be safer to do a check in the CMake code for realpath which would create a HAVE_REALPATH variable.

See Source/cmake/OptionsCommon.cmake for the other HAVE checks.

[Radar WebKit Bug Importer]

<rdar://problem/72062538>

[Fujii Hironori]

(In reply to Don Olmstead from comment #2)
> I feel like it'd be safer to do a check in the CMake code for realpath which
> would create a HAVE_REALPATH variable.

I think adding HAVE_REALPATH can't solve this problem.
I don't know which Unix WebKit is supporting, but Linux and FreeBSD can take NULL as the second argument of realpath, as far as I read the documents.
I think it's better to remove the old code.

[Yusuke Suzuki]

Comment on attachment 415526 [details]
Patch

Setting r- based on Hironori's comment. Could you update the patch :D ?

[Samuel Thibault]

Created attachment 415663 [details]
Patch

[Samuel Thibault]

I have done so

[Darin Adler]

This idiom occurs in other places, not just HyphenationLibHyphen.cpp. I found it in SandboxUtilities.mm, SandboxInitializationParametersCocoa.mm, AuxiliaryProcessMac.mm, FileSystemPOSIX.cpp, and StdFilesystem.cpp.

I think it’s peculiar to do this in only one place.

[Fujii Hironori]

Comment on attachment 415663 [details]
Patch

r- for comment#8.

[Samuel Thibault]

Apparently this fell in the cracks of my todo-list.

I have reworked the patch, submitted to https://github.com/WebKit/WebKit/pull/38913 , along other pull requests that fix building on GNU/Hurd (https://bugs.webkit.org/show_bug.cgi?id=285806)