Subject: | |
From: | |
Reply To: | |
Date: | Fri, 10 Mar 1995 15:00:26 PST |
Content-Type: | text/plain |
Parts/Attachments: |
|
|
Chris Bartram writes -
>
>Tell me it ain't so...
>
>C programs compiled on iX 5.0 WON'T execute on a 4.0 system??? What are the
>software developers out there doing to get around this? *is* there a way to
>compile on 5.0 so your programs will execute on previous operating systems???
>
Very timely, Chris. "WON'T execute" is a little harsh. "Unsupported"
and therefore "unrecommended" by HP is closer to home. It really
depends on what you are doing (i.e. what your application does) and
what libraries you are using. With a little tweeking, we were able
to get some vanilla apps compiled on 5.0 and successfully running on
both 4.0 and 5.0 (C.50.00).
HOWEVER, Jeff Woods discovered a quite nasty bug in the 5.0 C compiler.
There's a problem with address resolution of long pointers inside
structs; i.e. it's broke. So, if your program happens to run in
priv mode and setcritical and references long pointers in structs,
well, the results are less than satisfactory. (Jeff, details?)
So, to get around that we moved the (5.0) compiler and libraries to a
PUB50.SYS group, modified CCXL.PUB.SYS to check for a CC50 variable
and if set true, run the compiler and libraries out of PUB50. We
replaced the compiler and libraries from 4.0, so currently everything
now compiles on 5.0 using the 4.0 libraries and compiler.
After HP comes out with a patch for the (above) bug (sorry I don't
have an SR #), we'll have to re-think when we are going to bite
the bullit, start compiling on 5.0 and do regression testing on
4.0 - a little something I'll have to remember to put in my schedules
when the pre-powerpatch-pull-push release of 6.0 comes around...
--
_________________________________________________________________
Craig Vespe Unison Software
[log in to unmask] 811 Barton Springs, Austin Tx.
_________________________________________________________________
|
|
|