Re: Handicapped barcheck message in Lilypond version 2.17.26-1 (and also in Mac OSX PPC 2.16.2)

[David Kastrup]

Wim van Dommelen <address@hidden> writes:

> On 16 Sep 2013, at 21:12 , David Kastrup wrote:
>
>> Not before we get more data.  I don't see 0/0 in current master but
>> rather 1/2 as before.  So we need to get more data (which platform
>> exactly, which binary etc).  Since I can't reproduce this on current
>> master and there has not really been any relevant change I can
>> remember,
>> this points to a bad compilation and/or a platform-specific problem.
>> Or
>> something related to garbage collection.  We need to see whether we
>> can
>> reproduce this on another system.
>
> The problem is in the Mac OS X PPC binary. I downloaded and installed
> the file MacOS X PPC: LilyPond 2.17.26-1 (from
> http://www.lilypond.org/development.html).

> The Intel architecture runs OK, the PPC produces this strange
> error. Below the message from the PPC binary on the PPC system:

It's not going to make it easier to pinpoint the problem if we only get
to see it on PPC.  As I stated above "this points to a bad compilation
and/or a platform-specific problem.  Or something related to garbage
collection."

I'm not sure whether we use the same compiler version for the final
crosscompile on all GUB platforms.  If we do, any compiler bug would
likely be in the code generation backend rather than the front end.

> But then I went back to the last stable (2.16.2) release. And to my
> surpise it has the same problem......... So I was wrong and too quick
> in tying it to the new development version, this problem is older
> and/or deeper.
>
> I'm not sure how many people are still using the PPC binaries or how
> this binary is created, but apparantly there is some problem
> there. For the Mac-freaks: when I load the PPC binary on the Intel
> (and run it with the default Mac architecture emulation) exactly the
> same error message is produced which confirms to me it is that
> specific binary, not the architecture.

Well, how comfortable are you debugging binaries?  I can take a look at
the source code for any obvious "contenders", but being able to take a
good look at the actually operating code is going to go a long way
towards finding the problem.

-- 
David Kastrup