On 2 Dec 2004, at 20:43, Steve Smith wrote: >> Do you know which options trigger the DES breakage in RAT? I haven't >> been able to duplicate it on my test boxes... > > I've pretty-much always found it with any version of gcc > 2.95 under > Debian; however Rhys Hawkins once mentioned that he started seeing it > with when he added my alsa patch (and compiled with gcc3). Under > AMD64 nothing worked, and compiling qfDES with no compiler flags > caused gcc to segfault. There's probably a optimisation bug in gcc3, > but if qfDES can trigger it so consistently it implies to me that the > code is too fragile, and playing around edges of what is valid C. The > whole of my machine is compiled with the same compiler and it's > perfectly stable.
Interesting. I know some fixes went into the UCL tree for use with gcc3 after the AG-2.3 version of rat, but they don't change the qfDES code. It seems that something fixes the problems though - I've been running rat compiled with gcc3 for a months now (on MacOS X and FreeBSD 5) without problems. Colin
