[Unix] Adopt UnixFileDescriptor in IPC::Attachment

• Component: New Bugs
• Status: RESOLVED FIXED

[Zan Dobersek]

[Unix] Adopt UnixFileDescriptor in IPC::Attachment

[Zan Dobersek]

Created attachment 456697 [details]
Patch

[Carlos Garcia Campos]

Comment on attachment 456697 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=456697&action=review

It seems there's a caller missing in GTK.

> Source/WebKit/Platform/IPC/unix/ConnectionUnix.cpp:520
> +                if (!!attachments[i].fd()) {

I think it's worth adding Attachment::isValid() or isNull() or bool operator

[Zan Dobersek]

Created attachment 456796 [details]
Patch for landing

[Carlos Garcia Campos]

Comment on attachment 456796 [details]
Patch for landing

View in context: https://bugs.webkit.org/attachment.cgi?id=456796&action=review

> Source/WebKit/Shared/glib/UserMessage.cpp:94
> -            if (g_unix_fd_list_append(result.fileDescriptors.get(), attachment.releaseFileDescriptor(), nullptr) == -1)
> +            if (g_unix_fd_list_append(result.fileDescriptors.get(), attachment.fd().release(), nullptr) == -1)

Now that I see this again, I find a bit weird that the returned fd that it's still owned by the attachment can be released. Could we remove the non const getter and make attachment.release().release() instead? or maybe keep the releaseFileDescriptor() method in attachment that returns m_fd.release() instead?

[Zan Dobersek]

Created attachment 456801 [details]
Patch for landing

Less getters, more releasing.

[Zan Dobersek]

Created attachment 456806 [details]
Patch for landing

Fixed the reviewed-by line.

[Zan Dobersek (Reviews)]

Comment on attachment 456806 [details]
Patch for landing

Clearing flags on attachment: 456806

Committed r292458 (249309@trunk): <https://commits.webkit.org/249309@trunk>

[Zan Dobersek (Reviews)]

All reviewed patches have been landed.  Closing bug.

[Radar WebKit Bug Importer]

<rdar://problem/91346350>