|
 |
4416f5 |
From FEDORA_PATCHES Mon Sep 17 00:00:00 2001
|
|
|
4416f5 |
From: Kevin Buettner <kevinb@redhat.com>
|
|
|
4416f5 |
Date: Fri, 19 Mar 2021 11:07:11 -0700
|
|
|
4416f5 |
Subject: gdb-rhbz1941080-fix-gdbserver-hang.patch
|
|
|
4416f5 |
|
|
|
4416f5 |
;; Backport change which fixes gdbserver testing hang on f34 and rawhide.
|
|
|
4416f5 |
|
|
|
4416f5 |
Fix potential hang during gdbserver testing
|
|
|
4416f5 |
|
|
|
4416f5 |
We're currently seeing testing of native-extended-gdbserver hang while
|
|
|
4416f5 |
testing the x86_64 architecture on both Fedora 34 and Fedora Rawhide.
|
|
|
4416f5 |
The test responsible for the hang is gdb.threads/fork-plus-threads.exp.
|
|
|
4416f5 |
|
|
|
4416f5 |
While there is clearly a problem/bug with this test on F34 and
|
|
|
4416f5 |
Rawhide, it's also the case that testing should not hang. This commit
|
|
|
4416f5 |
prevents the hang by waiting with the "-nowait" flag in
|
|
|
4416f5 |
close_gdbserver.
|
|
|
4416f5 |
|
|
|
4416f5 |
The -nowait flag is also used in the kill_wait_spawned_process proc in
|
|
|
4416f5 |
gdb/testsuite/lib/gdb.exp, so there is precedent for doing this.
|
|
|
4416f5 |
|
|
|
4416f5 |
There are also 15 other uses of "wait -i" scattered throughout the
|
|
|
4416f5 |
test suite. While it's tempting to change these to also use the
|
|
|
4416f5 |
-nowait flag, I think it might be safer to defer doing so until we
|
|
|
4416f5 |
actually see a problem.
|
|
|
4416f5 |
|
|
|
4416f5 |
I've tested this patch on Fedora 32, 33, 34, and Rawhide. Results are
|
|
|
4416f5 |
comparable on Fedora 32 and 33. On Fedora 34 and Rawhide, with this
|
|
|
4416f5 |
commit in place, testing completes when the target_board is
|
|
|
4416f5 |
native-extended-gdbserver. On those OSes, when not using this commit,
|
|
|
4416f5 |
testing usually hangs due to a problem with
|
|
|
4416f5 |
gdb.threads/fork-plus-threads.exp. I've also tested on all of the
|
|
|
4416f5 |
mentioned OSes with target_board=native-gdbserver; for that testing,
|
|
|
4416f5 |
I achieved comparable results over a number of runs. (Unfortunately
|
|
|
4416f5 |
results are rarely identical due to racy tests.)
|
|
|
4416f5 |
|
|
|
4416f5 |
gdb/testsuite/ChangeLog:
|
|
|
4416f5 |
|
|
|
4416f5 |
* lib/gdbserver-support.exp (gdbserver_exit): Use the
|
|
|
4416f5 |
"-nowait" flag when waiting for gdbserver to exit.
|
|
|
4416f5 |
|
|
|
4416f5 |
diff --git a/gdb/testsuite/lib/gdbserver-support.exp b/gdb/testsuite/lib/gdbserver-support.exp
|
|
|
4416f5 |
--- a/gdb/testsuite/lib/gdbserver-support.exp
|
|
|
4416f5 |
+++ b/gdb/testsuite/lib/gdbserver-support.exp
|
|
|
4416f5 |
@@ -418,7 +418,12 @@ proc close_gdbserver {} {
|
|
|
4416f5 |
verbose "Quitting GDBserver"
|
|
|
4416f5 |
|
|
|
4416f5 |
catch "close -i $server_spawn_id"
|
|
|
4416f5 |
- catch "wait -i $server_spawn_id"
|
|
|
4416f5 |
+
|
|
|
4416f5 |
+ # If gdbserver misbehaves, and ignores the close, waiting for it
|
|
|
4416f5 |
+ # without the -nowait flag will cause testing to hang. Passing
|
|
|
4416f5 |
+ # -nowait makes expect tell Tcl to wait for the process in the
|
|
|
4416f5 |
+ # background.
|
|
|
4416f5 |
+ catch "wait -nowait -i $server_spawn_id"
|
|
|
4416f5 |
unset server_spawn_id
|
|
|
4416f5 |
}
|
|
|
4416f5 |
|