Re: recommended test tool chain for solaris9?

[Tom Tromey <tromey at redhat dot com>]

Andreas> Funny fact is, right now I do not see it. But always when I
Andreas> do a parallel make for example. Or when I have multiple
Andreas> builds running.

Odd.

I really don't see how the code could fail the way it does:

foreach x $srcfiles {
    set args [libjava_read_xfail [file rootname $x].xfail]
    if {[file exists [file rootname $x].no-link]} {
...

This snippet covers all uses of "x".
I looked and there don't seem to be any troublesome uses of "upvar"
or "uplevel" in the called code.  Likewise variable traces.
So I don't see how "x" could be undefined :-(

What does `runtest --version' say?  For me:

fleche. runtest --version
WARNING: Couldn't find the global config file.
Expect version is	5.26
Tcl version is		8.3
Framework version is	1.4.2.x

I don't know whether this is the currently blessed version.  It is
just what I got out of the src repo the last time I updated and
rebuilt gdb.


Are you using the same version across all platforms?  If so, maybe
there is some Tcl bug we're somehow triggering.  (Or maybe there is a
bug in our test harness -- but I don't see it.)


You could try something gross, like renaming "x" to something else.
If there is a stray upvar somewhere (e.g., somewhere deep in dejagnu),
this would actually help.

Tom