Apache OpenOffice (AOO) Bugzilla – Issue 8387
Please adapt fp exception macros for Linux to latest glibc interface
Last modified: 2003-03-13 11:10:03 UTC
Hi, the glibc provides now a common interface to the fp exceptions in fenv.h. Could you adapt the code as proposed in the attached patch. This makes porting easier to other Linux platforms. Gerhard
Created attachment 3188 [details] makes fp exception macros common for all Linux platforms
want opinion from porting project
Created attachment 4128 [details] Alternative FP_EXCEPTIONS implementation for Linux for S/390.
In case you don't want to change the implementation for all Linux platforms, I have attached an alternative implementation for Linux for S/390 only.
Hi, solmath.hxx.diff defintely *NOT* approved. You have removed a workaround that is needed for building under gcc 3.1 on ppc Linux. Please do NOT submit patches removing any PPC Linux pieces without first asking me. I have been trying to get all of the patches to support PPC Linux into the tree for almost 2 years now and pieces of them keep disappearing. This disappearing act needs no help! ;-) Also what is the minimum glibc version required to support your unifaction effort? Unless these macros existed under glibc-2.1.3, we should probably not change them since some distributions on non-x86 still use that glibc version "officially". Your solmath-s390-2.diff *is* approved since it will impact no one else. Kevin
Patch committed to branch cws_srx644_ooo20030223.
approved.
Just to add my 2 cts.: Kevin is right, as long as we use "old" glibc versions we definitely need the old hacked macros too. NOTE that all tools/inc/solmath.hxx functionality was shifted down to sal/inc/sal/mathconf.h and sal/inc/rtl/math.{h,hxx} on branch cws_srx644_rtlmath, which is to be merged back to mws_srx644 soon, and the S390-patch will have to be reapplied then.
The patch has been committed, but not yet verified.
in cws_srx644_ooo20030223
As mentioned on the qa dev list on March 5th I will close all resolved <wontfix/duplicate/worksforme/invalid> issues. Please see this posting for details.