make build hangs

[chill]

Hi guys,

I'm trying to use OpenBSD on my laptop. I followed the faqs for upgrading to 4.4-stable. The kernel built and installed fine but I am trying to rebuild the userland and the initial configuration hangs at "checking whether build environment is sane..."

You guys have any ideas I can try out when I get home? Sorry if I didn't post enough info. I am using amd64 version with GENERIC.MP kernel.

[jggimi]

This is not enough information to debug. The message you see comes from a (likely GNU) configure script, which is used hundreds of times during a userland build.

Is the OS hung, or is it just the build in some sort of loop?
If the OS is still operating, perhaps top(1) can provide an indication of what process is running in a loop.

If the OS is still running can ctrl-C from the shell interrupt the build?

BTW, if the OS hangs, you may still be able to find out why, by forcing a kernel dump per sysctl ddb.console=1 and crash(8).

Is this a repeatable problem? Does it occur at the exact same spot each time, or is it different? Different failure points in the build indicate hardware trouble (RAM, heat, power).

The key to diagnosing build problems -- that don't involve hanging the entire system -- is the script(1) tool in combination with col(1).
e.g.:

# script
Script started, output file is typescript
# make build
..... <build failure ...>
# exit
Script done, output file is typescript
# col -b < typescript > session
# less session
... < examine output for clues....
  Hint: start at the end and work upward>

[chill]

During initial configuration of "make build" before anything has built. GCC stops at "Checking if build environment is sane..."

The most I have waited was 15 minutes and still nothing. I have tried 4 times and it just stops right at that same line. Yes a good old Ctl+C will kill the build process.

My hardware is ok, my laptop runs cool and I ran about a 6 hr memtest on my memory a couple months ago with no errors.

I think I am just going to try patching the kernel instead and save me the trouble of building all the userland stuff anyways.

Thanks though.

[ocicat]

Originally Posted by chill

During initial configuration of "make build" before anything has built. GCC stops at "Checking if build environment is sane..."

Again, not a lot of information has been provided. One thing you may want to check is that you have the proper source downloaded & installed. If you don't have source already installed, you will get the correct tree through the following command:

# cvs -d$CVSROOT checkout -rOPENBSD_4_4 -P src

...as detailed in Section 5.3.3 of the FAQ:

http://openbsd.org/faq/faq5.html#BldGetSrc

If you have source already installed, it may or may not be the correct version, & the above command will not correct the error. The above command also assumes you had OpenBSD 4.4 installed. If source was installed before issuing the above command, it would have to be from 4.4-release. A common error made by those new to compiling the system is to grab the source to -current while on a -release installation. Even if the version numbers match, this is an incapable mix. Weirdnesses may very well occur.

I would highly recommend rereading Sections 5.1 - 5.3.

[jggimi]

Originally Posted by chill

... I have tried 4 times and it just stops right at that same line...

This is indicative of a software, or configuration problem, then. Hardware problems usually manifest as random, non-repeating failures.

... Yes a good old Ctl+C will kill the build process...

Then you could, if you ever had the interest, go investigate what is hung/looping from another console, since the OS is still running.

...My hardware is ok, my laptop runs cool and I ran about a 6 hr memtest on my memory a couple months ago with no errors.

In this particular case, hardware may not be the culprit. But I'd like to make a point, here. It is true that memtest and its variants can prove if you have a hardware problem. Unfortunately, they cannot prove that you don't. In general, six hours may not be sufficient testing length for RAM testing. I recently replaced a laptop's SO DIMM which did not indicate errors until the 22nd hour of memory testing.

I also like CPU stress testing, which heats up CPUs to the peak of their operating range. A userland build is not a perfect CPU stress tester, but it often can show hardware problems. For OpenBSD, I like sysutils/stress. For Windows workstations, I like Prime95's torture test.

[chill]

ocicat - I already used the FAQs as a manual to do this. This is exactly what I did.

# cd /usr
# export CVSROOT=anoncvs@obsd.cec.mtu.edu:/cvs
# cvs -d$CVSROOT checkout -rOPENBSD_4_4 -P src ports xenocara
# cd /usr/src/sys/arch/amd64/conf
# config GENERIC.MP
# cd ../compile/GENERIC.MP
# make clean && make depend && make
# make install
# reboot
# rm -rf /usr/obj/*
# cd /usr/src
# make obj
# cd /usr/src/etc && env DESTDIR=/ make distrib-dirs
# cd /usr/src
# make build

Apart from this I am on a totally clean install of 4.4, I have not done or installed anything else.

Then you could, if you ever had the interest, go investigate what is hung/looping from another console, since the OS is still running.

jggimi - i see what i can do