Bundle unified sources more tightly in projects with deep directory structures
Created attachment 348203 [details] Patch
Created attachment 348211 [details] Patch
Created attachment 348212 [details] Patch
Comment on attachment 348212 [details] Patch View in context: https://bugs.webkit.org/attachment.cgi?id=348212&action=review > Source/WebKit/ChangeLog:50 > +2018-08-27 Tim Horton <timothy_horton@apple.com> double change log
Created attachment 348213 [details] Patch
Created attachment 348214 [details] Patch
Created attachment 348215 [details] Patch
Comment on attachment 348212 [details] Patch View in context: https://bugs.webkit.org/attachment.cgi?id=348212&action=review > Source/WebCore/page/AlternativeTextClient.h:-34 > -#define USE_DICTATION_ALTERNATIVES 1 There's a FIXME in WebPageProxy.h about this.
Created attachment 348222 [details] Patch
Created attachment 348228 [details] Patch
r=me
Created attachment 348233 [details] Patch
Created attachment 348234 [details] Patch
The WPE failure is super weird.
Created attachment 348238 [details] Patch
Created attachment 348245 [details] Patch
Created attachment 348248 [details] Patch
Created attachment 348250 [details] Patch
Created attachment 348252 [details] Patch
Created attachment 348254 [details] Patch
Created attachment 348259 [details] Patch
Seems like X11 headers are super dirty (at least, have a macro `None`). So I'm de-unifying some GTK things to keep it building.
Created attachment 348261 [details] Patch
Created attachment 348263 [details] Patch
Created attachment 348264 [details] Patch
(In reply to Tim Horton from comment #22) > Seems like X11 headers are super dirty (at least, have a macro `None`). So > I'm de-unifying some GTK things to keep it building. Yeah, that None's pretty infamous. :/ X11 is 80s technology, I'm afraid.
Does this commit lead to a good build speedup? If so, then OK, but otherwise perhaps it's not worth it? It's not surprising that unifying files from different directories is harder.
BTW, I'll try to help with the GTK and WPE builds later today or tomorrow.
(In reply to Michael Catanzaro from comment #27) > Does this commit lead to a good build speedup? If so, then OK, but otherwise > perhaps it's not worth it? It's not surprising that unifying files from > different directories is harder. In WebKit2 it’s quite sizable. In WebCore there’s impact but less huge. I think WPE is ok, and Gtk is clooooose.
(In reply to Tim Horton from comment #29) > (In reply to Michael Catanzaro from comment #27) > > Does this commit lead to a good build speedup? If so, then OK, but otherwise > > perhaps it's not worth it? It's not surprising that unifying files from > > different directories is harder. ALSO, many of these fixes are errors that have crept in since we turned on unified sources and would affect our ability to build singly, so they're good to fix regardless. > In WebKit2 it’s quite sizable. In WebCore there’s impact but less huge. > > I think WPE is ok, and Gtk is clooooose.
Created attachment 348311 [details] Patch
Created attachment 348314 [details] Patch
Created attachment 348324 [details] Patch
I'm going to give up until mcatanzaro has a chance to peek at GTK.
OK. Right now I am attempting to figure out which of 235407-235410 (inclusive) has busted the layout tests so badly for GTK. I've never had so much trouble before. This bug is next!
Oh, just from glancing at EWS, you busted Windows too. :P The Ruby process is crashing. Congrats! [NOTE] You may have encountered a bug in the Ruby interpreter or extension libraries. Bug reports are welcome. For details: http://www.ruby-lang.org/bugreport.html 0 [main] ruby 8204 cygwin_exception::open_stackdumpfile: Dumping stack trace to ruby.exe.stackdump CMake Error at Source/cmake/WebKitMacros.cmake:47 (message): generate-unified-source-bundles.rb exited with non-zero status, exiting Call Stack (most recent call first): Source/JavaScriptCore/CMakeLists.txt:1227 (WEBKIT_COMPUTE_SOURCES)
Oddly windows built fine a few patches ago.
(In reply to Michael Catanzaro from comment #35) > OK. Right now I am attempting to figure out which of 235407-235410 > (inclusive) has busted the layout tests so badly for GTK. I've never had so > much trouble before. Did I break it with https://trac.webkit.org/changeset/235408/webkit ?
(In reply to Simon Fraser (smfr) from comment #38) > Did I break it with https://trac.webkit.org/changeset/235408/webkit ? Probably either that one or r235409... I wanted to narrow it down to the right commit, for starters. Thanks for your interest. Let's discuss in bug #189009.
Um, I meant bug #189063. Tim, I'll be back tomorrow to try to get GTK building here.
(In reply to Michael Catanzaro from comment #40) > Um, I meant bug #189063. > > Tim, I'll be back tomorrow to try to get GTK building here. Um, tomorrow again, sorry!
Just kidding, tonight!
Created attachment 348474 [details] Patch
Created attachment 348522 [details] Patch
Taking a stab at making the ruby slightly worse to not crash on windows.
Created attachment 348525 [details] Patch
*** Bug 189370 has been marked as a duplicate of this bug. ***
<rdar://problem/44198126>
https://trac.webkit.org/changeset/235521/webkit