On Thursday 11 October 2001 10:34, you wrote:
> Joseph Fannin wrote:
> > The problems with 2.95.x compilers is another stumbling block --
> > though some documents may say these compilers are unsupported for Linux,
> > they are "unofficially" supported for 2.[4|5], as is, to a lesser extent,
> > 3.0.x. Anything that won't build (and work) with at least 2.95.x is
> > broken.
> That may be the current opinion, but the fact is that there are compiler
> bugs in those versions of gcc that miscompile some of our perfectly
> legitimate code. See my TAKE message from yesterday, for example.
After sending off my last mail I found myself wishing I'd written that
I've been running an -xfs kernel (with XFS on /) for over a month now that
was built with the 2.95.4-prerelease in Debian unstable, and have had
absolutely no problems for over a month now. 2.96 never worked for me.
I'm a bit out of my league, since I'm not capable myself of doing what
I'm suggesting, but I would guess that if anyone wants to look into the
problems with the 2.95.x compilers (2.96 is a probably a red herring), you
might want to take a look at 2.95.4.
Really, wherever the problem lies, it *will* keep XFS out of the official
"Bull in pure form is rare; there is usually some contamination by data."
-- William Graves Perry Jr.