[lfs-dev] Oddities while testing.

Bruce Dubbs bruce.dubbs at gmail.com
Fri Jun 2 20:04:30 PDT 2017


Ken Moffat wrote:
> I've been building a partial system so that I'm ready for the 2017
> texlive.  A large part of the eventual test will be all the deps for
> biber (something like 100+ perl modules, including those needed for
> testing each dependency) so I wanted to use perl-5.26 in case any are
> not yet up-to-date.
>
> I built a system on another machine using 2017-05-13 versions, so I
> figured I would stick with those versions and just update perl (and
> try the rpc "replacement" for BLFS - I've already documented that).
>
> Well, apart from the fun of the perl-5.26.0 build with static libs
> (thanks, Armin, for explaining the problem), I had two odd issues.
>
> I build in Xorg, with one (urxvt) term for the build.
>
> In perl, I run the tests in chroot with 'make test >some.log 2>&1'.
>
> I've had issues in the past with lib/Benchmark.t hanging (bad kernel
> config), so when I noticed the perl tests seemed to be taking a long
> time I opened another term and ran 'tail -f' on the log.  To my
> consternation, the only thing in the log was the output from building
> the test progs, nothing from the tests themselves.  Fortunately, the
> tests did complete (with the normal extra failures for that machine)
> and at that point the log showed all the results.
>
> But I'm sure I've run 'tail -f' on perl test logs in chroot on some
> previous builds without problem.
>
> The other odd thing was vim - I run the tests, and then I add '||
> true' so that the expected failure won't break my script by
> returning false.  But the tests got to there, and then nothing more
> happened, i.e. it did not start to install vim (it just sat there
> for hours).  I cancelled that, retried, again it hung.  So for the
> moment I've commented out the vim tests.  Unfortunately I overwrote
> the original vim testlog when I retried, but it looked like a normal
> completion.
>
> At the moment these are just two more data points for the continuing
> "ain't life weird?" saga.  Hope your builds are not so odd as mine
> :)

Interesting comments Ken.  I just did a fresh update today via jhalfs.  My 
initial attempt of perl in Chapter 5 didn't work, but I used Armin's patch 
(translated to a sed) and all the tests seemed pretty normal.  There were 
a couple of new failures in the overall lfs tests, but when I reran the 
tests in chroot but after Chapter 6 was completed, those new tests passed. 
  At that point I just documented the issue in the book and committed.

The util-linux tests may need some attention.  In the log is:

   "Executing the tests in parallel (12 jobs)"

The only real problem with that is that it throws off the timing data a 
bit.  The total build/test time of 42 seconds is pretty quick. :)

As far as blfs tests go, I've started using a construct like

(make -k check || true) &&
...

I find that running the tests in a subshell seems to encapsulate the tests 
and handle the return code better.

I'll also note that I do not have any issues with nfs using the current 
packages/instructions in BLFS using gcc7 based lfs systems, but security 
is not a big factor as I am the only one using my internal network and it 
is protected fairly well from the outside.

   -- Bruce




More information about the lfs-dev mailing list