| Summary: | [GTK][WPE] CFLAGS from pkg-config for (E)GL are not passed to WebKit | ||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Product: | WebKit | Reporter: | Adrian Perez <aperez> | ||||||||||
| Component: | WebKitGTK | Assignee: | Carlos Alberto Lopez Perez <clopez> | ||||||||||
| Status: | RESOLVED FIXED | ||||||||||||
| Severity: | Normal | CC: | aestes, beidson, bugs-noreply, cgarcia, clopez, commit-queue, darin, mcatanzaro, youennf, ysuzuki, zan | ||||||||||
| Priority: | P2 | ||||||||||||
| Version: | Other | ||||||||||||
| Hardware: | All | ||||||||||||
| OS: | Linux | ||||||||||||
| See Also: |
https://bugs.webkit.org/show_bug.cgi?id=145408 https://bugs.webkit.org/show_bug.cgi?id=178081 |
||||||||||||
| Attachments: |
|
||||||||||||
Carlos López has posted a working patch at http://sprunge.us/aUhc Created attachment 317384 [details]
Patch
Comment on attachment 317384 [details]
Patch
CMake is so fragile. How could this happen. :(
I mean, the fact that it has been possible to build at all, for years, without the right flags being passed... that is surprising. Comment on attachment 317384 [details] Patch Clearing flags on attachment: 317384 Committed r220326: <http://trac.webkit.org/changeset/220326> All reviewed patches have been landed. Closing bug. (In reply to Michael Catanzaro from comment #4) > I mean, the fact that it has been possible to build at all, for years, > without the right flags being passed... that is surprising. Two things explains this: * The default build (with Mesa as GL library and with X11 support built into Mesa's EGL) works fine without passing any cflag. You only run into this failure if your GL library requires some cflag to build. Requiring this is the exception rather than the norm. I have only seen this with the Freescale GL libraries (Vivante) or when you disable the X11 support when building Mesa's EGL libraries (as Adrian comments). * When we landed r184954 <https://trac.webkit.org/r184954>, we didn't used any (E)GL function on the UIProcess, so it was not needed to add this there at that time. Only in WebCore was needed by then. Committed r220331: <http://trac.webkit.org/changeset/220331> Comment on attachment 317384 [details] Patch View in context: https://bugs.webkit.org/attachment.cgi?id=317384&action=review > Source/WebKit/CMakeLists.txt:97 > + if (USE_OPENGL) This doesn't account for libepoxy, where OPENGL and EGL packages aren't leveraged. > Source/WebKit/CMakeLists.txt:127 > list(APPEND WebKit2_INCLUDE_DIRECTORIES > "${THIRDPARTY_DIR}/ANGLE" > "${THIRDPARTY_DIR}/ANGLE/include/KHR" We shouldn't be using ANGLE headers in the first place. Comment on attachment 317384 [details] Patch View in context: https://bugs.webkit.org/attachment.cgi?id=317384&action=review >> Source/WebKit/CMakeLists.txt:127 >> "${THIRDPARTY_DIR}/ANGLE/include/KHR" > > We shouldn't be using ANGLE headers in the first place. ... but it's apparently needed due to indirect GraphicsContext3D.h inclusions. This should be addressed later then, after GC3D usage cleanup. (In reply to Zan Dobersek from comment #9) > Comment on attachment 317384 [details] > Patch > > View in context: > https://bugs.webkit.org/attachment.cgi?id=317384&action=review > > > Source/WebKit/CMakeLists.txt:97 > > + if (USE_OPENGL) > > This doesn't account for libepoxy, where OPENGL and EGL packages aren't > leveraged. > Good catch. I didn't realized about libepoxy. Let's re-open. > > Source/WebKit/CMakeLists.txt:127 > > list(APPEND WebKit2_INCLUDE_DIRECTORIES > > "${THIRDPARTY_DIR}/ANGLE" > > "${THIRDPARTY_DIR}/ANGLE/include/KHR" > > We shouldn't be using ANGLE headers in the first place. That was already there before, I didn't introduced this include of the ANGLE headers. Reopen Created attachment 317412 [details]
Patch
Created attachment 317413 [details]
Patch
Created attachment 317414 [details]
Patch
fix typo on changelog
Comment on attachment 317412 [details]
Patch
The ChangeLog is messed up and all EWS are red....
(In reply to Michael Catanzaro from comment #16) > Comment on attachment 317412 [details] > Patch > > The ChangeLog is messed up and all EWS are red.... Why do you comment on an obsoleted attachment? please see https://bugs.webkit.org/attachment.cgi?id=317414&action=review Comment on attachment 317414 [details]
Patch
I'm bad at Bugzilla! Much better.
Comment on attachment 317414 [details] Patch Clearing flags on attachment: 317414 Committed r220334: <http://trac.webkit.org/changeset/220334> All reviewed patches have been landed. Closing bug. |
For example, in “FindEGL.cmake”, the “Cflags” field from “egl.pc” is stored in the “${EGL_DEFINITIONS}” CMake variable, which is used to pass the flags to WebCore, but the value is never passed to WebKit, and building WebKit can fail. One such case is when building in a Wayland-only system (no X11 support), where Mesa's “egl.pc” contains “-DMESA_EGL_NO_X11_HEADERS”, but not passing that when building WebKit causes Mesa's “eglplatform.h” to try to include the X11 headers — and the build fails.