topic Re: Interesting PASCAL failure in Operating System - OpenVMS

Interesting PASCAL failure

Brian Reiter — Wed, 06 Dec 2006 04:57:44 GMT

Hi folks,

I thought I'd float this question here, see if I get a response while embarking on the tortorous route of getting our reseller to organise the ticket.

The pascal compiler (V5.9-95) crashed when building a debug version of an image. (Stack trace attached). This is unusual as the non-debug version build quite happily, in the past I've has problems with optimised code versus non-optimised code.

Any clues while I wait for HP to get in touch?

regards

Brian

Re: Interesting PASCAL failure

Brian Reiter — Wed, 06 Dec 2006 05:25:42 GMT

Hmmm,

And it fails on the v6 compiler as well. (see attached).

cheers

Brian

Re: Interesting PASCAL failure

Volker Halle — Wed, 06 Dec 2006 05:31:19 GMT

Brian,

the error seems to be happening in the same routine in both compiler versions: DBG_FIND_LOCAL_THUNK

The routine name seems to imply, that it's being used when compiling /DEBUG.

Volker.

Re: Interesting PASCAL failure

John Gillings — Wed, 06 Dec 2006 17:15:58 GMT

Brian,

By definition, any compiler generated stack dump is a bug which should be formally reported to HP.

If you can provide a (small?) reproducer, the compiler team will usually respond with a fix quickly. As long as it's not too unwieldy, the reproducer may then be added to the compiler regression test suite.

Re: Interesting PASCAL failure

Brian Reiter — Thu, 07 Dec 2006 02:59:42 GMT

John,

I've already put that in motion (through the reseller as we don't have a support contract).

I'll have a word with the engineer and see if we can put to together a reproducer, can't guarantee a small one though.

cheers

Brian

Re: Interesting PASCAL failure

John Reagan — Thu, 07 Dec 2006 08:55:07 GMT

Drat. I thought I fixed all of those for V6.0.

As you guessed, this is with debug support for schema types (even STRING). V5.9 had some problems, but I spent a day or two cleaning it up for V6.0.

I would need to see a reproducer (I'll take something big as long as I have all the files needed to compile the program).

While you are waiting for the "official" response, feel free to send the sources directly to me at "john dot reagan at hp dot com"

John
Pascal Project Leader

Re: Interesting PASCAL failure

Brian Reiter — Thu, 07 Dec 2006 09:46:00 GMT

Hi John,

Thanks for that. The engineer is working on the reproducer at the moment. It should be with you soon (I hope).

cheers

Brian