If I run "Tools/Scripts/run-javascriptcore-tests --gtk" on my Linux box, it will run build-jsc with the --32-bit argument, which is not what I want, given that I have already built JSC for x86-64. The patch to be attached fixes this issue. (If I omit --gtk, even having already built JSC, I get worse results, but that's another bug.)
Created attachment 113320 [details] Patch
Comment on attachment 113320 [details] Patch View in context: https://bugs.webkit.org/attachment.cgi?id=113320&action=review > Tools/Scripts/webkitdirs.pm:250 > + if (isGtk() or isEfl()) { > + determineConfigurationProductDir(); > + my $host_triple = `grep -E '^host = ' $configurationProductDir/GNUmakefile`; I think EFL uses CMake so this check on the GNUMakefile will likely always fail for them. Maybe remove the isEfl() test?
Created attachment 114080 [details] Patch
Updated patch. R? :)
Comment on attachment 114080 [details] Patch View in context: https://bugs.webkit.org/attachment.cgi?id=114080&action=review > Tools/Scripts/webkitdirs.pm:250 > + determineConfigurationProductDir(); > + my $host_triple = `grep -E '^host = ' $configurationProductDir/GNUmakefile`; > + $architecture = $1 if $host_triple =~ m/^host = ([^-]+)-/; In the Mac case they first check the architecture set for the build and then if it's not there, they fall back to the system architecture. I think you should do the same here.
Created attachment 114228 [details] Patch
Comment on attachment 114228 [details] Patch View in context: https://bugs.webkit.org/attachment.cgi?id=114228&action=review > Tools/Scripts/webkitdirs.pm:254 > + # Fall back to output of `arch', if it is present. Ultra nitpick, you don't really have further fallbacks, so I guess the "if it is present" is a bit misleading. You'll use that or fail in some way I suppose. Anyway, this looks OK to me.
Comment on attachment 114228 [details] Patch Clearing flags on attachment: 114228 Committed r99729: <http://trac.webkit.org/changeset/99729>
All reviewed patches have been landed. Closing bug.