This is the mail archive of the
gdb-patches@sourceware.org
mailing list for the GDB project.
Re: [RFA/hpux] acknowledging TTEVT_LWP_EXIT
- From: Mark Kettenis <mark dot kettenis at xs4all dot nl>
- To: guitton at adacore dot com
- Cc: gdb-patches at sourceware dot org
- Date: Tue, 18 Sep 2007 12:20:20 +0200 (CEST)
- Subject: Re: [RFA/hpux] acknowledging TTEVT_LWP_EXIT
- References: <20070918094806.GA1859@adacore.com>
> Date: Tue, 18 Sep 2007 11:48:06 +0200
> From: Jerome Guitton <guitton@adacore.com>
>
> In inf-ttrace, when a TTEVT_LWP_EXIT event is received (thread
> exiting), the corresponding thread is removed from the thread list
> and never resumed. Which is wrong ; the ttrace man page reads:
>
> TTEVT_LWP_EXIT This event flag indicates that the debugger wants to
> be notified when a thread is exiting via the
> lwp_exit() system call. The thread stops upon entry
> to the system call.
>
> So the dying thread is stopped, it should be resumed one last time.
> Otherwise, any other thread waiting for its death on pthread_join
> would be blocked forever (e.g. in attachment, a simple program which
> freezes when it is run under GDB).
>
> Also in attachement, a patch that fixes this issue. It also fixes 3
> unexpected failures of the testsuite on HPUX 11.11.
>
> 2007-09-18 Jerome Guitton <guitton@adacore.com>
>
> * inf-ttrace.c (inf_ttrace_private_thread_info): New structure type.
> (inf_ttrace_delete_dying_threads_callback): New function.
> (inf_ttrace_resume): After resuming the execution, iterate over
> the dying threads to delete them for the thread list.
> (inf_ttrace_wait): on TTEVT_LWP_EXIT and TTEVT_LWP_TERMINATE,
> mark the corresponding thread as dying instead of removing it
> from the thread list.
> (inf_ttrace_thread_alive): return 0 for dying threads.
Looks ok to me.