develooper Front page | perl.perl5.porters | Postings from October 2016

Re: Intermittent failures in 'make' at "Updating 'mktables.lst'"

Thread Previous | Thread Next
Karl Williamson
October 1, 2016 04:11
Re: Intermittent failures in 'make' at "Updating 'mktables.lst'"
Message ID:
On 09/30/2016 02:03 PM, James E Keenan wrote:
> On 09/08/2016 11:43 AM, Karl Williamson wrote:
>> On 09/04/2016 01:51 PM, James E Keenan wrote:
>>> Occasionally, when building blead or a branch directly derived from
>>> blead, I experience a failure during 'make' at this point:
>>> #####
>>> Writing tables
>>> Making pod file
>>> Making test script
>>> Updating 'mktables.lst'
>>> 1 error
>>> make: stopped in /usr/home/jkeenan/gitwork/perl
>>> #####
>>> To simply be told that there was an error and that 'make' therefore
>>> stopped is not very informative.
>>> The first time this happened -- the two times I can recall were both on
>>> a FreeBSD 10.3 VM -- I then called 'make' and something that was clearly
>>> a syntax error eventually emerged.
>>> Today, when I called 'make' after the error message above, 'make'
>>> resumed and completed successfully -- i.e., no error and no indication
>>> why I 'make' stopped the first time through.
>>> I haven't been able to reproduce this error.  Has anyone else
>>> experienced this?
>> As far as I recollect, mktables itself should always try to output an
>> error message if it itself discovers an error.  I have experienced
>> intermittent failures in writing the tables, but none with the exact
>> symptoms you describe.  This used to happen a lot, actually, on one of
>> Tux's machines.  He and I were trying to debug it together, getting
>> closer and closer, and suddenly it stopped occurring.  We pieced
>> together that what had happened was that he had rebased, and had picked
>> up a commit that fixed the problem.  It was a Dave Mitchell commit, and
>> the bug it fixed was in the I/O portion of the interpreter.  I do see
>> something very occasionally since then, so it is some race in the I/O, I
>> guess.  I haven't looked for this in a while, but it did appear that
>> often when this happened, the file had been slightly
>> scrambled, but usually the garbage was in a comment at the beginning and
>> had no effect on the file's operation.  I have had it fail in writing
>> mktables.lst too, and don't remember now who I got past it, except to
>> retry.
>> If this is happening with any regularity, you could try running it under
>> valgrind (do this in the background so you don't have to wait seemingly
>> forever), and see what happens.
>>> Thank you very much.
>>> Jim Keenan
> khw: ISTR that we had some discussion subsequent to these posts,
> probably on #p5p, about this 'mktables.lst' problem.  IASTR testing
> repeated builds on my FreeBSD-10.3 VM, but beyond that I don't recall
> the upshot of that discussion.
> In any event, today I went to smoke-test petdance's numeric.c changes on
> my FreeBSD-10.3 VM, and I got the problem again:
> #####
> Writing tables
> Making pod file
> Making test script
> Updating 'mktables.lst'
> 7 errors
> make: stopped in /usr/home/jkeenan/gitwork/perl
> #####
> I then called:
> #####
> make -j8 test_harness
> #####
> ... and 'make' and 'make test_harness' went on their merry way without
> incident.
> Thank you very much.
> Jim Keenan

I don't recall any further conversation, nor do I have any more info 
than what you quoted above.  When you did 'make' again, did it recompile 
mktables?  I wonder what '7 errors' means

Thread Previous | Thread Next Perl Programming lists via nntp and http.
Comments to Ask Bjørn Hansen at | Group listing | About