johnh@wheaton.UUCP (03/01/87)
We have discovered a problem (we think) with the Powell compiler from
Dec under unix. (We have discovered and fixed several others).
The problem is if you have a definition module x which exports procedure
z and in implementation module x you have a 'inline' module y which exports
procedure z the symbol which is created for procedure z is x_y_z. If you
have another module which imports procedure z from module x the linker
complains that symbol x_z is undefined. This is even if the keyword
@unqualified is put before procedure z inside module y.
A work around is to declare procedure z in the main module x and have that
call y.z with z being exported qualified.
Is there a better workaround?
We have fixed several problems associated with returns in 'fuction' procedures.
We also have an idea on how to get the system to complain about def files which
have been modified since an object module was compiled at link time.
Any suggestions, comments or additional bugs fixed or not would be appreciated.
johnh...
--
-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
UUCP: ihnp4!wheaton!johnh telephone: (312) 260-3871 (office)
Mail: John Hayward Math/Computer Science Dept. Wheaton College Wheaton Il 60187
Act justly, love mercy and walk humbly with your God. Micah 6:8bken@rochester.UUCP (03/01/87)
Yes, the Powell compiler doesn't handle nested modules correctly. You can also get into trouble with mutually recursive references in imported modules even when there is no real circularity. This is due to the way the definitions are handled (by textual inclusion). Would someone official from DECWRL care to say whether it is legal to post patches for bugs? I'm sure we could all get a better compiler by sharing our fixes. Ken
maner@andy.bgsu.edu.UUCP (03/01/87)
I suppose you sent the bug fixes on the Powell at DECWRL. When you did, did you get any impression about when he'll get around to a new release with all bug patches remedied? Thanks for the answer. Walter Maner BEST CSNet maner@research.bgsu.edu : UUCP ... cbosgd \ : !osu-eddie!bgsuvax!maner : ... cbatt / : Office (419) 372-8719 IF Tuesday THEN live ELSE machine : Secretary (419) 372-2337 WORST Mail CS Dept, BGSU, Bowling Green, OH 43403 "It's not what you know, it's Who you know."
ken@CS.ROCHESTER.EDU.UUCP (03/01/87)
My impression is that Powell is no longer doing any work on the compiler as distributed to universities. But I don't really know any better. I would be glad to hear any news from DECWRL. Guys? Ken