This is the mail archive of the
libc-alpha@sourceware.org
mailing list for the glibc project.
Re: [PATCH] Run ld.so tests only if $(build-shared) is yes
- From: Roland McGrath <roland at hack dot frob dot com>
- To: "Joseph S. Myers" <joseph at codesourcery dot com>
- Cc: "H.J. Lu" <hjl dot tools at gmail dot com>, GNU C Library <libc-alpha at sourceware dot org>
- Date: Tue, 9 Oct 2012 14:16:33 -0700 (PDT)
- Subject: Re: [PATCH] Run ld.so tests only if $(build-shared) is yes
- References: <20120910130629.GA20970@intel.com><20120910230816.49F592C0A2@topped-with-meat.com><Pine.LNX.4.64.1209102355080.886@digraph.polyomino.org.uk>
> There seem to be several properties one might wish to assign to tests.
> shared-only is one, but so is whether they go in tests or xtests, and so
> is whether they are built and run by standard rules or need their own
> rules. Then there's distinguishing tests using locales to set LOCPATH for
> them (I'm not sure why we don't just set LOCPATH for all tests, but if
> there's a reason to avoid that then having a flag on the relevant tests
> would seem better than hardcoding the relevant LOCPATH setting in many
> places).
>
> Rather than adding lots of variables for the various combinations of these
> properties, maybe there should be a way of setting a list of properties
> for a test - whether in a variable for that test, or through having the
> "tests" variable setting contain things such as
> "test-name:property-1:property-2" rather than just plain test names?
>
> (If a test is shared-only, and you're generating PASS / FAIL lines for
> tests, arguably there should be an UNSUPPORTED line for the test that
> didn't run because you had a static build.)
These are all reasonable high-level ideas. I agree that something more
coherent and formal would be better than the status quo and the additional
hair it is accreting. Someone needs to propose the exact details and offer
to do the implementation work, and then we can hash it all out.
Thanks,
Roland