Bug 21219

Summary:	RFE: classify architectures where enhanced features are not build for
Product:	[Retired] Red Hat Raw Hide	Reporter:	Enrico Scholz <rh-bugzilla>
Component:	glibc	Assignee:	Jakub Jelinek <jakub>
Status:	CLOSED ERRATA	QA Contact:	Aaron Brown <abrown>
Severity:	medium	Docs Contact:
Priority:	medium
Version:	1.0	CC:	fweimer
Target Milestone:	---	Keywords:	FutureFeature
Target Release:	---
Hardware:	i386
OS:	Linux
Whiteboard:
Fixed In Version:		Doc Type:	Enhancement
Doc Text:		Story Points:	---
Clone Of:		Environment:
Last Closed:	2000-11-22 18:56:14 UTC	Type:	---
Regression:	---	Mount Type:	---
Documentation:	---	CRM:
Verified Versions:		Category:	---
oVirt Team:	---	RHEL 7.3 requirements from Atomic Host:
Cloudforms Team:	---	Target Upstream Version:
Embargoed:

Description Enrico Scholz 2000-11-22 02:38:57 UTC

At some places in glibc.spec you are checking e.g. for

%ifnarch sparcv9 i586 i686 athlon alphaev6

to disable certain parts. For people wanting to build all parts with an
optimal target, it's very annoying to search for all places such constructs
are appearing.

I would prefer the creation of a macro like

%define unwanted_archs	 sparcv9 i586 i686 athlon alphaev6

at the top of the .spec-file and its usage instead of the `%ifnarch ...'
above.
Then a full-build can be enabled easily be commenting out the
macro-definition.

Comment 1 Jakub Jelinek 2000-11-22 17:43:26 UTC

IMHO you don't want to build them with anything other than the base archs,
because that just means you're no longer able to build statically linked
base arch packages (i386, alpha, sparc). And who cares about optimized
profiling library when the actual slow down caused by profiling hides any
gains.

Comment 2 Enrico Scholz 2000-11-22 18:56:11 UTC

I have no need to keep compatibility with i386 (local machines are i586 and
above; to provide binaries distributed widely, there are existing special
machines). Why should I use slower settings when faster ones are existing?

I am not interested in *optimized* profiling libs, but I want profiling (&
devel) libs and I do not see any reason to compile glibc twice (once for i686,
once for i386 to get glibc-profile).

I don't want that you enable building for i686 generally, but allowing to decide
it at a *one* place  would make life easier for people (like me) who are
building their whole system for --target=i686.

Comment 3 Jakub Jelinek 2000-12-18 22:51:54 UTC

See glibc-2.2-9.