Fix PR cli/15603
authorTom Tromey <tromey@redhat.com>
Tue, 18 Jun 2013 19:57:49 +0000 (19:57 +0000)
committerTom Tromey <tromey@redhat.com>
Tue, 18 Jun 2013 19:57:49 +0000 (19:57 +0000)
commit427cd150eed8c0dd4f0d0a1105448b4ebe36da6d
treeed94cd3b70e5de75895c7abcb57bc78ad1f83911
parentc31f39360b3865a248055587778301e85da250fc
Fix PR cli/15603

This fixes PR cli/15603.

The bug here is that when a software watchpoint is being used, gdb
will stop responding to C-c.  This is a regression caused by the
"catch signal" patch.

The problem is that software watchpoints always end up on the bpstat
list.  However, this makes bpstat_explains_signal return
BPSTAT_SIGNAL_HIDE, causing infrun to think that the signal is not a
"random signal".

The fix is to change bpstat_explains_signal to handle this better.  I
chose to do it in a "clean API" way, by passing the signal value to
bpstat_explains_signal and then adding an explains_signal method for
watchpoints, which handles the specifics.

Built and regtested on x86-64 Fedora 18.
New test case included.

* break-catch-sig.c (signal_catchpoint_explains_signal): Add 'sig'
argument.
* breakpoint.c (bpstat_explains_signal): Add 'sig' argument.
Special case signals other than GDB_SIGNAL_TRAP.
(explains_signal_watchpoint): New function.
(base_breakpoint_explains_signal): Add 'sig' argument.
(initialize_breakpoint_ops): Set 'explains_signal' method for
watchpoints.
* breakpoint.h (struct breakpoint_ops) <explains_signal>: Add
signal argument.
(bpstat_explains_signal): Likewise.
* infrun.c (handle_syscall_event, handle_inferior_event): Update.

* gdb.base/random-signal.c: New file.
* gdb.base/random-signal.exp: New file.
gdb/ChangeLog
gdb/break-catch-sig.c
gdb/breakpoint.c
gdb/breakpoint.h
gdb/infrun.c
gdb/testsuite/ChangeLog
gdb/testsuite/gdb.base/random-signal.c [new file with mode: 0644]
gdb/testsuite/gdb.base/random-signal.exp [new file with mode: 0644]