On Mon, 2005-05-23 at 17:25 +0100, Richard Earnshaw wrote: > On Mon, 2005-05-23 at 17:20, Jeffrey A Law wrote: > > On Mon, 2005-05-23 at 16:34 +0100, Richard Earnshaw wrote: > > > On Mon, 2005-05-23 at 16:19, Eric Christopher wrote: > > > > > > http://gcc.gnu.org/wiki/GCC%204.1%20Slush > > > > > > > > > > Just to let folks know that mips-elf fails to build newlib. > > > > > There's a segfault in is_gimple_reg_type(), which is being > > > > > passed a null type. I'm not sure if I'll have time to look > > > > > at it tonight. > > > > > > > > I took a look and it seemed to work for me, I'll double check with a > > > > clean tree and all multilibs in a bit. > > > > > > > > -eric > > > > > > This will be PR 21638. There's been several proposed patches for this > > > now, but none seems to have been committed. It is getting a bit silly. > > It's not a question of getting a bit silly, but a question of being > > thorough. Daniel's checking patch has exposed several more cases were > > we are not getting those flag bits set correctly and IMHO this situation > > has shown how important it is to get those bits set correctly. > > > > I'd much rather take the time to fix all these problems, install the > > fixes along with the checking bits to ensure they never come back > > rather than to iterate on each one separately. > > > > However, in the mean-time I'm stuck. I can't build my world anymore, so > I can't test the compiler.... I understand. But realize that I'm trying to ultimately save you time in the long run -- in the mean time you can pretty easily hack around these problems.
> Maybe we should start a timer on this one. I would recommend against. Every one of these missing calls to compute the status bits properly is a latent bug waiting to be exposed. I'd much rather bite the bullet and get them fixed now. The fact that it's affecting a lot of people keep the coals hot on my feet :-) jeff
