There is a slight bug in the RegExp JIT code where we use the m_containsNestedSubpatterns before the JIT compilation has set it. The result is that this limits somewhat the regexps that can take advantage of JIT'ed greedy sub patterns.
Created attachment 347944 [details] Patch
Comment on attachment 347944 [details] Patch View in context: https://bugs.webkit.org/attachment.cgi?id=347944&action=review r=me > Source/JavaScriptCore/yarr/YarrJIT.cpp:3565 > + codeBlock.setUsesPaternContextBuffer(); nit: "Patern" looks mis-spelled. Maybe fix in a follow up patch since it's not related to this issue.
Comment on attachment 347944 [details] Patch View in context: https://bugs.webkit.org/attachment.cgi?id=347944&action=review >> Source/JavaScriptCore/yarr/YarrJIT.cpp:3565 >> + codeBlock.setUsesPaternContextBuffer(); > > nit: "Patern" looks mis-spelled. Maybe fix in a follow up patch since it's not related to this issue. I fixed that in the patch I'm working on. I can factor that out if you like.
(In reply to Michael Saboff from comment #3) > Comment on attachment 347944 [details] > Patch > > View in context: > https://bugs.webkit.org/attachment.cgi?id=347944&action=review > > >> Source/JavaScriptCore/yarr/YarrJIT.cpp:3565 > >> + codeBlock.setUsesPaternContextBuffer(); > > > > nit: "Patern" looks mis-spelled. Maybe fix in a follow up patch since it's not related to this issue. > > I fixed that in the patch I'm working on. I can factor that out if you like. Either works for me.
Comment on attachment 347944 [details] Patch Clearing flags on attachment: 347944 Committed r235238: <https://trac.webkit.org/changeset/235238>
All reviewed patches have been landed. Closing bug.
<rdar://problem/43654564>