[Clfs-dev] More issues with trunk

[Ken Moffat]

2009/10/1 Chris Staub <chris at beaker67.com>:
> On 10/01/2009 12:37 PM, Ken Moffat wrote:
>>
>>  From my x86_64-64 build, which is now building its kernel:
>>
>> 2. e2fsprogs' configure fails because pkg-config is not installed.
>> The easy option is to just add it, as in LFS.
>
> This is already dealt with in the book by using PKG_CONFIG=true, though I
> agree it would be easier just to add pkg-config to the book.
>

 Hmm, that variable didn't jump out at me when I was looking to see
what had changed.

>> 3. I was unable to remove the coreutils directory after running
>> its testsuite, and counldn't umount the filesystem (in use).  I
>> happened to have lsof on the host, that said it couldn't stat
>> a particular test file which it thought was a tmpfs (although it
>> was within the coreutils tests directory) and labelled '(deleted).
>> After remounting r/o during shutdown, and then rebooting, I
>> was able to remove it.  Might be down to how I build, or might
>> turn out to be a common problem.
>
> Someone on the LFS lists reported the same problem a while back. I
> investigated and found that the Coreutils testsuite does mount something in
> its source dir, though it should also be unmounted as well. I have not
> figured out why it would not be unmounted - presumably the only reason for
> that would be that the testsuite (or that particular test) didn't finish. I
> did run across the problem once myself, though that was while I was testing
> dependencies and had deliberately left out some programs to see what would
> happen.
>
> http://linuxfromscratch.org/pipermail/lfs-dev/2009-March/062643.html
>

 (see below - I missed the chown - that's what comes from
taking my eye off the ball)

>> 4. The iputils man-pages are not in doc/ : the patch creates
>> them in the top-level directory, so change
>> install -v -m644 doc/*.8 /usr/share/man/man8
>> to
>> install -v -m644 *.8 /usr/share/man/man8
>
> They are in doc/* for me. Did you use -Np0 or -Np1 on that patch?
>

 -p1 as for all patches in the book..  Odd,   Now I've looked at the
patch, I agree they get
created in doc/, but they weren't there when the last part of
the install failed, and repeating it with fresh source apparently
installed them successfully as '*.8'.

 Oh, I see your point. :-(  Who decided to mix -p0 and -p1
patches together ?

>> 5. In general, I see no point in commenting on testsuites
>> unless the book is near to a release, but I think the
>> following are worth noting:
>>
>> 5.1 coreutils failed in the gnulib tests -
>>    CC     read-file.o
>> read-file.c:136: fatal error: opening dependency file
>> .deps/read-file.Tpo: Permission denied
>> compilation terminated.
>> make[5]: *** [read-file.o] Error 1
>
> This should be covered by the chown command in the Coreutils instructions.
>
 Oh.  Another detail change which I missed.

 Thanks for taking the time to reply.

ĸen
-- 
After tragedy, and farce, "OMG poneys!"