This is the mail archive of the libc-alpha@sourceware.org mailing list for the glibc project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

[PATCH] Use base-machine in abilist tuple.


When generating the sparc abilist entries I found that we use
the config-machine for the tuple.  This makes the regexps on
sparc quite non-trivial.

We use base-machine in other contexts to generate ABI files, for
example libc-abis.h, so it seemed appropriate to use it for generating
the check-abi-config tuple as well.

This means I can use two sparc abilist regexps:

	sparc-.*-linux.*/thread
	sparc64-.*-linux.*/thread

and that takes care of all of the wacky machine names we can
config for:

	sparcv8
	sparcv9
	sparcv9b
	sparcv9v
	sparcv9v2
	sparc64b
	sparc64v
	sparc64v2

and so on and so forth.

I considered using abi-name, if set, but that would make the
sparc regexps have a different format than the other targets.
In such a setup abi-name based regexps would be of the form:

	MACHINE-OS.*

instead of:

	MACHINE-.*-OS.*

which handles the VENDOR part which isn't included in abi-name.

This should have no impact on any targets with existing abilist
entries.  I ran a sanity check-abi on x86-64 just to make sure,
and that went fine.

Any objections?

	* Makerules (check-abi-config): Use base-machine instead of
	config-machine.

diff --git a/Makerules b/Makerules
index 18aeb9e..3531bbe 100644
--- a/Makerules
+++ b/Makerules
@@ -1176,7 +1176,7 @@ endif
 ifeq ($(firstword $(sysd-sorted-done) f)$(firstword $(generating) f),tf)
 config-tls := thread
 check-abi-config := \
-  $(config-machine)-$(config-vendor)-$(config-os)/$(config-tls)
+  $(base-machine)-$(config-vendor)-$(config-os)/$(config-tls)
 endif
 
 update-abi-%: $(..)scripts/merge-abilist.awk $(..)abilist/%.abilist \


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]