Tests with keys that match both PASS, FAIL (or now
optionally XPASS), count as fail. XPASSes were previously
ignored. Handling them as FAIL seems the most useful
alternative, but not counting XPASSes may be deliberate.
It's also a matter of compatibility, so make it optional.
Attempts to use --handle-xpass-as-fail was previously
flagged as a usage error. If you pass it now, on state with
previous mixed XPASS and PASS results but doesn't change in
this run, the XPASS is discovered as a (new) regression.
For new XPASSing tests, it's handled as a new FAIL.
* btest-gcc.sh (--handle-xpass-as-fail): New option.
This is a long-standing bug: passing "-j --add-passes-despite-regression"
or "--add-passes-despite-regression -j" caused the second option to be
treated as TARGET; the first non-option parameter.
* btest-gcc.sh (Option handling): Handle multiple options.
I use objs-gcc.sh as a preparatory step before calling
btest-gcc.sh in my scripts, for example my cris-elf
autotester. I thought, why not use it for native builds
too. Except that use, with binutils release-style tarballs
and a x86_64-pc-linux-gnu host, was broken. Now that I look
at it, the script seems to have aged poorly... Still,
there's a need for such a script to install stuff needed for
btest-gcc.sh (and to fix up stuff if needed), and this can
still be that script. So, I prefer to fix show-stoppers for
common uses, while taking care to retain compatibility for
use that could possibly still work, with current sources.
A long time ago (before 2011, but after this script was
created in 2002, and used for a few years), the binutils
(and gdb and gcc) toplevel Makefile may have had a bootstrap
target that worked with binutils but didn't require gcc
sources to be present. Now, you'll get an error (see
configure.ac line 1366 and on). Let's just build the
default make-target when "bootstrap" is known to fail.
An alternative would be to fold this native
non-i686-pc-linux-gnu clause into the native
i686-pc-linux-gnu clause, as that seems to have been
originally intended as *the* single native clause, but
that'd require further edits (e.g. to remove install-dejagnu
and make gdb build conditional on gdb sources presence, to
work with binutils tarballs, and I'd also then prefer to
build not just ld, but also gas and binutils).
As it's a minimal obvious change required for current native
use with release-tarballs and git-checkout use(*), I'm
installing this as obvious.
*) Native i686-pc-linux-gnu remains broken for other use
than specially constructed combined trees where dejagnu is
included at the toplevel (i.e. historic Cygnus devo-type).
contrib/regression:
* objs-gcc.sh: Only bootstrap if source-directory contains gcc.
* GCC_Regression_Tester.wdgt/widget.html: Mark as HTML 4.01,
which it now is, rather than XHTML, which it never was.
(<head>): Mark as UTF-8. Add title. Move CSS to top of
document, specify type. Mark script as 'defer'.
(updateContents): If loaded from HTTP, look for status in same
place as widget.
(gotContents): Use DOM methods to change text rather than innerHTML.
(<body>): Eliminate unnecessary DIV element.
* GCC_Regression_Tester.wdgt/Info.plist: Update version, copyright
notice.
From-SVN: r131145
2006-01-18 Andrew Pinski <pinskia@physics.uc.edu>
* btest-gcc.sh: gcc.sum has moved to gcc/testsuite/gcc/gcc.sum.
g++.sum has moved to gcc/testsuite/g++/g++.sum.
objc.sum has moved to gcc/testsuite/objc/objc.sum.
From-SVN: r109937
