Mail Archives: cygwin/2009/04/05/23:21:33
Hi all,
[ re: ... don't update! ]
Or at least, not without taking a backup of your Cygwin installation
first, at any rate the system-related parts (/bin /usr/lib /usr/include and
who knows if maybe more?), at least until we figure out whether everyone else
is going to have the same utter nightmare it has caused for me, or whether
it's just some quirk in my local environment. It might be ok if the ability
to compile stuff from source is not important to you, it might even be
something that's only wrong on my machine and won't affect you, but I'd feel
lousy if anyone else has to go through this and I could have warned them.
Everything was going fine. I've been building and patching and testing and
rebuilding binutils CVS constantly over the past couple of days and my new
patch was working fine.
Then around seven pm this evening I updated my installation, getting just
the latest DLL and the new M4 package.
Immediately after that, I tried to build binutils CVS from clean HEAD
sources. When I tested the new tools, the ld 'bootstrap' tests had all
regressed; I get error messages (in $objdir/ld/ld.log) that read:
tmpdir/ld1:../bfd/.libs/libbfd.a: file format not recognized; treating as
linker script
tmpdir/ld1:../bfd/.libs/libbfd.a:1: syntax error
After seven^Weight^Wnine (and counting, I've been gradually writing this
email while trying to fix it and the figure keeps getting out of date) solid
hours of doing nothing but rolling packages backwards and forwards and
rebuilding (or attempting to) binutils and the DLL dozens of times each, I am
still nowhere near recovering a working Cygwin installation.
My installation was working before, but it was a custom build of the DLL
that I can no longer reproduce, because although I still have the DLL itself I
no longer have the headers and libs to go with it, and have updated my sandbox
since.
Even if I knew the exact time/date to check out a CVS version ... I couldn't
rebuild it because I have nothing that currently works. All the DLL binary
packages in the distro either have the old speclib bug (which breaks attempts
to build things), or now that that's fixed, have the latest changes that also
appear to break everything. I say "appear to" because I can't do anything to
find out; I can't build anything, so I can't debug anything. Neither old nor
new versions of the DLL can be successfully used to attempt to compile an
intermediate version of the DLL, or of binutils. (There are so many
permutations of old-or-new compiler, old-or-new-binutils, old-or-new-dll,
old-or-new contents in /usr/include and /usr/lib that it's possible one of
them will work, but it could be a long time before I find it.)
So, in case we can avoid this happening to anyone else:
Q1) Has anyone out there updated since 4pm-ish (GMT+1) this afternoon, and
has a checkout of CVS binutils they could try building, and then testing with
make check-ld RUNTESTFLAGS="bootstrap.exp"
Q2) Has anyone out there /not/ updated today, and also has a checkout of CVS
binutils they could try building and testing?
In either case:-
If you get six fails, that's probably the bug I have here.
If you get one fail (in the "bootstrap with --static" test), you most likely
have the speclib generation bug.
(If you have either of these problems, you cannot rebuild the cygwin DLL,
meaning you are limited to the available downloads... all of which manifest
either one or the other problem.)
If you can successfully build binutils HEAD, and you get all passes in the
boostrap tests, can you please send me your "cygcheck -cd" output so I can
figure out what I have to install to get working again?
Meanwhile, I'm going to see if I can get it working statically with gcc3,
then attempt to cross-boostrap my 1.7 using my old 1.5; I'll report back if
any of that helps.
cheers,
DaveK
--
Unsubscribe info: http://cygwin.com/ml/#unsubscribe-simple
Problem reports: http://cygwin.com/problems.html
Documentation: http://cygwin.com/docs.html
FAQ: http://cygwin.com/faq/
- Raw text -