|Summary:||Repatch assembly dump: allow having no CodeBlock|
|Product:||WebKit||Reporter:||JF Bastien <jfbastien>|
|Severity:||Normal||CC:||commit-queue, fpizlo, jfbastien, keith_miller, mark.lam, msaboff, sbarati|
|Version:||WebKit Nightly Build|
|Bug Depends on:|
Description JF Bastien 2017-02-21 14:02:17 PST
Repatch's assembly dump assumes there's a CodeBlock present. WebAssembly doesn't have one. It crashes.
Comment 2 Filip Pizlo 2017-02-21 14:12:06 PST
Comment on attachment 302304 [details] patch I think you should make it easier to say this, since we say it in multiple places. In fact, the usual dumping style for CodeBlock+CodeOrigin is: dataLog(pointerDump(codeBlock), " ", codeOrigin) I think that's what we want here as well. But maybe we can take it a step further and create a helper for it. The right way to go is something like: dataLog(FullCodeOrigin(codeBock, codeOrigin)) Where FullCodeOrigin is a simple value class that holds a CodeBlock* and a CodeOrigin, and has a `void dump(PrintStream&) const` that prints them out.
Comment 3 JF Bastien 2017-02-21 16:41:47 PST
Created attachment 302337 [details] patch Add a helper class as Fil suggested, and use it in a bunch of places. This is nice.
Comment 4 WebKit Commit Bot 2017-02-21 16:42:59 PST
Comment 5 JF Bastien 2017-02-21 16:47:37 PST
Comment on attachment 302337 [details] patch Style fix.
Comment 7 WebKit Commit Bot 2017-02-21 17:18:28 PST
Comment on attachment 302339 [details] patch Clearing flags on attachment: 302339 Committed r212782: <http://trac.webkit.org/changeset/212782>
Comment 8 WebKit Commit Bot 2017-02-21 17:18:32 PST
All reviewed patches have been landed. Closing bug.