[Info-vax] Some questions on software for VMS 7.3 VAX

[hb]

On 01/26/2016 10:35 AM, lists at openmailbox.org wrote:
> On Tue, 26 Jan 2016 09:49:54 +0100
> hb via Info-vax <info-vax at rbnsn.com> wrote:
>>>>> Anybody wants to try the workaround, anyway?
>>> Sure if you want to spend the time messing around with it I'll gladly
>>> test it and get back to you. Especially the gcc idea sounds promising
>>> because the optimizations seem essential and that is less likely to
>>> break badly.

The corruption seems to be a wrong SP caused by inlining a function
which in turn uses alloca() to make room for a filename on the stack. So
the SP is roughly off by the length of that name. Unfortunately I can't
reproduce it with a small test program. But as far as I understand the
machine listing, the problem shows there.

Using /nooptimize turns inlining off and turns the problem down. I
changed the source to use #pragma noinline for that function and that
avoids the problem as well - and still optimizes everything else; how
good is now a different question.

The compiler is probably the latest available for VAX: Compaq C V6.4-005.

I made a new binary kit: https://www.sendspace.com/file/2qdd4v

The zip file is compressed with -1 for faster processing, so it is about
10% bigger than the first kit. I patched the GNU_STARTUP.COM to define
BLD_ROOT relative to where GNU_STARTUP.COM is: if that works as expected
you no longer need to define (or adjust the definition of) BLD_ROOT. The
old version of this command procedure is still there.

I'm sure you let me know if there are problems :-)