bootstrap/10028: gcc 3.1 and later up to 3.2.2 included, does not boostrap on alpha Tru64 Unix 5.1B using the native Compaq C compiler
pagnin@sssup.it
pagnin@sssup.it
Tue Mar 11 15:26:00 GMT 2003
>Number: 10028
>Category: bootstrap
>Synopsis: gcc 3.1 and later up to 3.2.2 included, does not boostrap on alpha Tru64 Unix 5.1B using the native Compaq C compiler
>Confidential: no
>Severity: serious
>Priority: medium
>Responsible: unassigned
>State: open
>Class: sw-bug
>Submitter-Id: net
>Arrival-Date: Tue Mar 11 15:26:00 UTC 2003
>Closed-Date:
>Last-Modified:
>Originator: Davide Pagnin
>Release: gcc version 3.2.2
>Organization:
>Environment:
uname -a
OSF1 V5.1 2650 alpha
CC = /bin/cc (native Compaq C compiler)
>Description:
In the 3.1 and subsequent series of gcc, a workaround for a fstat (and friends) class of function not properly exposed in Tru64 Unix has been introduced.
(Relevant PR ID number is: 6126)
The workaround, that applies to Tru64 Unix 5.1A (and to 5.1 as far as I know) is included in the os and platform specific file:
gcc/config/alpha/osf.h
as a define for CPP_SUBTARGET_SPEC:
-D__EXTERN_PREFIX
This workaround plays bad with fixes that Compaq (now HP) has introduced to Tru64 Unix 5.1B, and in particular with the new /usr/include/sys/stat.h file
By issuing a make bootstrap, after building the first stage, you get errors indicating that _F64__F64_ symbols aren't defined (this is due to the fact that both HP fixes and gcc osf workaround help in add a _F64_ symbol before *stat class of function and ending up in add that symbol twice!)
By removing -D__EXTERN_PREFIX the compilation completes cleanly (but then the old bug will return if you compile the source gcc on a Tru64 Unix 5.1A system)
>How-To-Repeat:
On a Tru64 Unix 5.1B alpha system:
setenv CC /bin/cc
./configure
make bootstrap
>Fix:
remove -D__EXTERN_PREFIX on 5.1B osf
perhaps a proper workaround can be generated by correctly fix the /usr/include/sys/stat.h file, through a fixinclude macro (tough I have not the knowledge do produce it)
>Release-Note:
>Audit-Trail:
>Unformatted:
More information about the Gcc-prs
mailing list