Bug 154317 – Modern IDB: More WK2 IPC Scaffolding

• Status: RESOLVED
• Resolution: FIXED
• Component: WebCore Misc.

[Brady Eidson]

Modern IDB: More WK2 IPC Scaffolding

This includes new messages.in files, all the messages that will be needed, and declarations for all the encoders/decoders, etc.

Very little of it is actually implemented for now.

[Brady Eidson]

Created attachment 271506 [details]
Patch v1

[Brady Eidson]

Seems like the Linux CMake builds *should* be finding the headers that are being complained about... not sure why they aren't.

[Michael Catanzaro]

Hm, if I replace #include <WebCore/IDBCursorInfo.h> with #include "IDBCursorInfo.h" in the DerivedSources file, it makes it to the next #include... investigating.

[Michael Catanzaro]

I'm not making any progress on this tonight, sorry; will try again tomorrow if someone doesn't figure it out first.

I have not managed to figure out how it is the #include <WebCore/foo.h> syntax works at all for any foo.h that does not live in a directory named WebCore in some include path. I've grown accustomed to this style of #includes, but never before stopped to think about how it works... seems like magic now....

[Martin Robinson]

(In reply to comment #4)
> I'm not making any progress on this tonight, sorry; will try again tomorrow
> if someone doesn't figure it out first.
> 
> I have not managed to figure out how it is the #include <WebCore/foo.h>
> syntax works at all for any foo.h that does not live in a directory named
> WebCore in some include path. I've grown accustomed to this style of
> #includes, but never before stopped to think about how it works... seems
> like magic now....

These forwarding headers used to be generated by a CMake macro called WEBKIT_CREATE_FORWARDING_HEADERS into the WebKitBuild/Release/DerivedSources/ForwardingHeaders/ directory. Each forwarding header is composed of a single #include of the path to the real file. Now I don't see that called anywhere for WebKitGTK+, unless I'm missing it.

[Carlos Garcia Campos]

Forwarding headers are created by the generate-forwarding-headers.pl that scans the given directory looking for headers referenced that need a forwarding header. The problem here is that the header that requires the forwarding header is only referenced by generated code, and the script only scans Source/WebKit2. In all other cases it works because there's code in Source/WebKit that also has the same header, so when the derived sources are compiled, the forwarding header has already been created. So, to fix this, we should explicitly include <WebCore/IDBCursorInfo.h> in WebIDBConnectionToServer.cpp instead of relying on it being included by the messages header. Another solution would be to add another target in the makefile to run generate-forwarding-headers.pl right after the ipc code is generated by before it's compiled to create forwarding headers for the WebKit2 derived sources dir.

[Brady Eidson]

(In reply to comment #6)
> Forwarding headers are created by the generate-forwarding-headers.pl that
> scans the given directory looking for headers referenced that need a
> forwarding header. The problem here is that the header that requires the
> forwarding header is only referenced by generated code, and the script only
> scans Source/WebKit2. In all other cases it works because there's code in
> Source/WebKit that also has the same header, so when the derived sources are
> compiled, the forwarding header has already been created. So, to fix this,
> we should explicitly include <WebCore/IDBCursorInfo.h> in
> WebIDBConnectionToServer.cpp instead of relying on it being included by the
> messages header. Another solution would be to add another target in the
> makefile to run generate-forwarding-headers.pl right after the ipc code is
> generated by before it's compiled to create forwarding headers for the
> WebKit2 derived sources dir.

I'll include it for now - It *will* have to be included eventually, anyways.

Thanks for figuring this out!

[Brady Eidson]

Created attachment 271554 [details]
Patch for landing

[Brady Eidson]

Created attachment 271557 [details]
Patch for landing v2

[Brady Eidson]

Created attachment 271560 [details]
Patch for landing v3

[Brady Eidson]

Created attachment 271561 [details]
Patch for landing v4

[Brady Eidson]

This is never ending...

[Brady Eidson]

Well it appears that the optimizations that clang does to realize methods aren't needed don't exist here.  More implementation here we come...

[Brady Eidson]

Created attachment 271564 [details]
Patch for landing v5

[WebKit Commit Bot]

Comment on attachment 271564 [details]
Patch for landing v5

Clearing flags on attachment: 271564

Committed r196705: <http://trac.webkit.org/changeset/196705>