Alexander, I'm working on it. The only thing I can say right now, that it's -ftree-vrp that broke the code -- building with -O -ftree-vrp makes broken binary, while -O alone does not. (sorry for top-posting) -----Original Message----- From: "Alexander Kabaev"<kabaev_at_gmail.com> Sent: 26.05.07 21:58:38 To: "Yuriy Tsibizov"<Yuriy.Tsibizov_at_gfk.ru> Cc: "ed_at_fxq.nl"<ed_at_fxq.nl>, "current_at_freebsd.org"<current_at_freebsd.org>, "freebsd_at_unixfreunde.de"<freebsd_at_unixfreunde.de> Subject: Re: [Regression] snd_emu10k1 doesn't work after GCC 4.2 upgrade On Sat, 26 May 2007 21:45:10 +0400 "Yuriy Tsibizov" <Yuriy.Tsibizov_at_gfk.ru> wrote: > I've re-checked this problem today and it seems to be a real gcc > regression. I had more pessimistic optimisation flags ( -O ) in my > first tests and snd_emu10k1 was running without problems. If I build > it with -O2 (default value for -CURRENT, with __MAKE_CONF=/dev/null) > it is broken (I checked it on my Audigy card and it shows different > problems from Live!, because it uses different code paths in driver). > snd_emu10kx is not broken with both -O and -O2. I can't tell for > shure, but there was no bug reports about broken snd_emu10k1 with -O2 > and old gcc. > > Yuriy > Show me the broken code. It is not that hard to isolate one if you know that it exists. Above rambling does not cut it. -- Alexander KabaevReceived on Sat May 26 2007 - 17:38:58 UTC
This archive was generated by hypermail 2.4.0 : Wed May 19 2021 - 11:39:11 UTC