7.0-STABLE -> 7.1-PRERELEASE Problems

[DutchDaemon]

Hi,

Today I decided to upgrade my FreeBSD 7.0-STABLE (about 2 months old) to FreeBSD 7.1-PRERELEASE.

The whole build/install process goes well, but I end up with a non-working C compiler, and it has now happened twice in a row (I had to borrow another BSD 7 system's C compiler for the second try).

The affected (non-functioning) files appear to be

/usr/bin/cc
/usr/libexec/cc1
/usr/libexec/cc1obj
/usr/libexec/cc1plus
/usr/bin/as
/usr/bin/ld

(replacing these with another system's binaries gives me a working C compiler anyway)

Has anyone else tried to compile and run FreeBSD 7.1-PRERELEASE (the standard 32-bits version from csup) today or _very_ recently and encountered problems like these? The error cc produces on execution is 'Abort Trap 6', by the way.

Note: I did clear /usr/obj between the first build and the second, but not a make clean in /usr/src. Trying that now for the third build.

[DutchDaemon]

The third attempt was successful. I saw an updated src/Makefile creep in during a csup. It may have something to do with it. I also cleared out the CPUTYPE setting from /etc/make.conf, though I hardly think that was the problem, since it's been in there for over a hundred builds on dozens of machines, ever since the pentium4 target was added. Anyhoo, back in business. Thanks for reading

[hamba]

This bit me as well yesterday and now my email server is down

I hope to get it up soon, I used sysinstall to see if I could revert back to 7.0-RELEASE but I lost all access to the server after a reboot.

I hope that it didn't mess everything up to much.

[Weaseal]

Originally Posted by hamba

This bit me as well yesterday and now my email server is down

This is precisely why -STABLE is not recommended for production environments. It's fun to play around with new features at home, but when it counts, stick to -RELEASE.

[DutchDaemon]

When I rebooted for the first time (after the first make installworld) I lost (remote) access to the server as well; it pinged, but no ports were open. When I had physical access to the machine I saw it hung right after an ntpdate command (during the loading of daemons). A simple Ctl+C set the wheels in motion again. It didn't happen a second time, though. The interesting thing was that all other binary stuff (/bin, /usr/bin, /usr/sbin, etc.) worked. Only cc and its associated helpers were actually in disrepair.

[hamba]

All I have is remote access, I'm in London and the server is in Berlin.

I got my ISP on it so hopefully they can get me up and running soon.

[phoenix]

When running -STABLE, always be sure to subscribe to the stable@freebsd.org mailing list. There was a big HEAD'S UP about this on the list yesterday. Everything should be working today, though.

[DutchDaemon]

Oh, my head's up now, P.

[hamba]

True but I've been running stable on my servers for as long as I can remember and this was the first time I had such a big failure. I finally am now able to get back into my system and I'm rebuilding everything.

[DutchDaemon]

Me too. Been running STABLE since 2.2.5, and maybe had three problems in 1000+ installs and updates (usually every 2/3 months for every install).

[Weaseal]

Originally Posted by hamba

True but I've been running stable on my servers for as long as I can remember and this was the first time I had such a big failure.

Originally Posted by DutchDaemon

Me too. Been running STABLE since 2.2.5, and maybe had three problems in 1000+ installs and updates (usually every 2/3 months for every install).

Certainly I agree that in a vast majority if your experience, -STABLE will almost never nuke your system, and when it does, you can get it back online quickly. I always run it on my desktop, and occasionally even bump around in -CURRENT.

However, if you are using it in a work/production environment, where 99.95% uptime is generally expected (and that .05% is always announced days if not weeks in advance), those unexpected outages could mean some tense moments between you and your boss.

[phoenix]

Ah, but in situations where you need 99.95% uptime, you have a testing environment setup where every little update gets tested before it hits the production servers, so there's not much risk to running -STABLE in that setup.

Bottom line to running -STABLE (or -CURRENT): subscribe to the relevant mailing list, and always stay a couple days behind the major MFCs and other imports.