Re: testcase for "absolute source" patch

[Michael Chastain <mec dot gnu at mindspring dot com>]

The first time I run openp.exp, I get 4 non-PASS results:

  FAIL: gdb.base/openp.exp: openp s1-da (non-existent source, basename, find in <dir>/$cdir/basename)
  FAIL: gdb.base/openp.exp: openp s3-da (non-existent source, relative name, find in <dir>/$cdir/relative)
  FAIL: gdb.base/openp.exp: openp s4-da (non-existent source, in the current dir, find in <dir>/$cdir/basename)
  FAIL: gdb.base/openp.exp: openp s5-da (non-existent source, in the upper dir, find in <dir>/$cdir/relative)

The second time I run openp.exp, it crashes and burns:

  Running /berman/fsf/_current_/source/gdb/HEAD/src/gdb/testsuite/gdb.base/openp.exp ...
  gdb compile failed, gcc: /tmp/migbat-testgdb-BqdGwxc7/test/gdb.base/openp/subdir/src/openp.c: No such file or directory
  WARNING: Testcase compile failed.

  ERROR: (timeout) GDB never initialized after 10 seconds.
  UNRESOLVED: gdb.base/openp.exp: openp b1+ (existing binary, basename)
  ...

I suspect this line is at fault:

  remote_exec host "mv $srcdir/$subdir/$testname.c $srcfile"

$srcdir is read-only.

More feedback:

. add a ChangeLog entry

. add some code at the top to check for build != host

    if { [is_remote host] } {
      unresolved "This test script does not work on a remote host."
      return -1
    }

. since the script does not work with build != host, you can simplify
  all the "remote_exec host" calls to TCL primitives that run on the
  build machine anyways.  Like: "file mkdir ...".  Your choice.
  If you like "remote_exec host", you can keep it.

. error checking on every external call.  Sometimes directory creation
  does fail.

That's all the time I have for this round.  I still haven't really
gotten into the substantive part of the script yet.  :(

Michael