|
From: Julian S. <js...@ac...> - 2011-10-28 02:35:40
|
Stephen, > Seeing this patch go by reminds me that I'd noticed a similar > seemingly incorrect set of constrains for the versions of the add.w > and sub.w instructions that take another register and an optional > shift rather than an immediate. I'm still working on an appropriate > test suite addition (it's slow going since I only have software > emulators). > > If no-one besides me has run into it yet, I was guessing that my > change wasn't important enough to rush into 3.7, but I'm appending it > below in case anyone wants to take a look; it will be in Bugzilla in > the future. Um, I wasn't aware this patch existed. What bugzilla number tracks this? My (permanent) concern is adding support for instructions without test cases, since it's exceedingly difficult to track down subtle errors without test cases. See https://bugs.kde.org/show_bug.cgi?id=245925#c3 for a sorry tale. So I've taken to fixing any such case that turns up in the wild, or has a regtest to exercise it. Anyway .. if you have a patch + test case then I will be very happy to push that into trunk, + into 3_7_BRANCH if anybody reports it as a failure in practice. J |