From: Philipp K. K. <pk...@sp...> - 2012-05-06 12:06:26
|
On 04.05.2012 20:22, Erik Petrich wrote: > > On Thu, 3 May 2012, Philipp Klaus Krause wrote: > >> I do not see any regression failures and do not know of any bugs >> in the optralloc-hc08 branch. There are a few things related to >> the work done there that IMO should be done, but I think they >> could (and probably should) be done after merging the >> optralloc-hc08 branch: >> >> * Right operand for GET_VALUE_AT_ADDRESS. * Review peephole rules >> (the new allocator results in different code being generated, so >> some old peephole rules may be unnecessary, while some new >> peephole rules will be desirable). * Make --stack-auto the >> default. >> >> Erik, what is your opinion on merging the optralloc-hc08 branch? >> Are there any bugs (e.g. visible only on some architectures or >> with certain command-line options) that need to be resolved >> before the merge? >> >> Philipp > > It seems to be working properly, both as-is and with --stack-auto > as the default. While I can also think of a few more improvements, > it's nothing drastic that should prevent merging back to trunk > now. > > Erik >From the snapshot regression results, in seems to SIGSEGV crash on Solaris, both sparc and x86. Philipp |