Re: [Nmh-workers] OpenBSD added to the buildbot cluster

[David Levine]

Ralph wrote:

> Does anyone run the test suite using valgrind(1)?  :-)

Yes, actually :-)  One of the tests (an important one, on inc)
relies completely on valgrind; it is skipped if valgrind is not
available.  Another test can be told to use valgrind.

I manually run valgrind sporadically and have picked up and
fixed some problems.  Clearly, more would be better.

I completely agree that our efforts would be better spent in
migrating to higher level abstractions.

And I'll note that nmh (MH, actually, though we might have added
some of them more recently) already provides some (low-level)
convenience functions.  But they're not used consistently.  If I
was to write nmh from scratch, I'd wrap every system/C lib call
so that we could easily deal with things like this.  But not
everyone agrees with that approach.  Anyway, again, I'd rather
see high-level abstractions.


Ingo wrote:

> Actually, David Levine already agreed in private mail that
> an audit would make sense

Well, my exact words were:

    Maybe we'll get to that task some day.  In the meantime,
    the significant noise from the linker could obscure more
    pressing problems.  Therefore, I have suppressed it from
    our build summary, though it's still in the build logs.

I would argue that the OpenBSD linker should have an option
to disable the warnings, but unfortunately I don't have the
time or motivation to pursue that.

David