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

[lists at openmailbox.org]

I thought I responded to this but I don't see my post. Sorry if this is a
duplicate.

On Mon, 25 Jan 2016 23:36:23 +0100
hb via Info-vax <info-vax at rbnsn.com> wrote:

> On 01/25/2016 09:42 PM, Stephen Hoffman wrote:
> > It's far more likely that the OpenVMS VAX and C tools have limitations
> > and omissions here.   The emacs code is very likely in good shape. 
> > Emacs is used all over the place, after all.
> 
> The VMS code of emacs very likely is not used very often and therefore
> may not be in the same good shape as the other emacs code.

And the same good shape as the other Emacs code is only imaginary and then
only applies to Linux. Most developers these days don't seem to be aware
that anything besides Linux and Windows exist.

> Anyway, there was a stack corruption. Compiled with /debug and /noopt the
> problem doesn't show.

I'll refrain from speculating whether the VAX/VMS C compiler is broken or
whether this is because of gcc-isms and other tricks the Emacs developers
played with the C standard and non-compliant, non-portable code that
actually compiles but is not guaranteed to work anywhere except under gcc.
I would bet a keg of beer on it being the latter. Oh darn I didn't refrain
after all.

> A workaround may be to compile one module, DIRED, with /noopt. Yes, I
> know, this makes EMACS even slower :-) To really fix the problem one
> would need to analyze and understand the corruption. So far I didn't
> spend the time.

I thought you said you got it to crash by control-X space. If so this isn't
limited to dired. Maybe the whole thing needs to be compiled /noopt and
then it will probably be unusable. The other option if you really want to
compile it like it was "designed" to be compiled if there indeed is any
such thing would be to get the comtemporaneous port of gcc for VAX/VMS and
build with that. gnu crapware is not guaranteed to build or work with
anything but gcc and bash and then only sometimes.

> 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.

-- 
Please do not copy me on mailing list replies. I read the mailing list.
RSA 4096 fingerprint 7940 3F02 16D3 AFEE F2F8  ACAA 557C 4B36 98E4 4D49