Re: Make define-builtin-markup{, -list}-command #:category #:properties keywords (issue160048)

[David Kastrup]

Neil Puttock <address@hidden> writes:

> 2009/12/3 David Kastrup <address@hidden>:
>
>> I keep asking for the testing, with little success so far.  I still
>> don't know whether the changes from a week ago solve the memory
>> leak/corruption problem reported with a previous version.
>
> Thanks, the latest patchset works fine (see attached test results),
> and the annoying `Parsed object should be dead' warnings have gone.

Good.  Very good.

I think that there may be a _functional_ difference in comparison to the
original approach: I suspect you now need to define your markups at top
file level with regard to modules, whereas previously it was ok to do it
anywhere in a subordinate module/scope as well.

But I think that this is not a serious drawback.

That means that I can go ahead seriously with cleanup and documentation
work.

>> I might have mentioned that I have not been able to verify the
>> original report, since it worked and I got no recipe to make it fail
>> in the reported way.
>
> I can't give you a recipe to make it fail on your system.  I mentioned
> what caused the problems on my system, but I can't give you any more
> info than I've already provided.
>
> To recap:
>
> I build the lilypond binary:
>
> ./autogen.sh --disable-optimising

That's what I have used as well.

> make -j2
> make install

Apart from -j2, same here.  I have a single processor machine.  Maybe
the problem occurs only when multithreading?  Is Lilypond multithreaded?

> Regression testing:
>
> make test-clean
> make -j2 CPU_COUNT=2 test-baseline
>
> apply patch
>
> make -j2 CPU_COUNT=2 check
>
> The previous patch which failed only worked without -j2 on my system,
> but spat out memory leaks even on blank files.

Wow.  Nothing here.  I have to admit that I _do_ get two undead smob
warnings right after writing out internals.texi.

However, I get them with or without my patch.

-- 
David Kastrup