This is the mail archive of the
gdb-patches@sourceware.org
mailing list for the GDB project.
Recent tracepoint regressions
- From: Keith Seitz <keiths at redhat dot com>
- To: "gp >> \"gdb-patches at sourceware dot org ml\"" <gdb-patches at sourceware dot org>
- Cc: Tom Tromey <tromey at redhat dot com>
- Date: Tue, 23 Apr 2013 14:12:48 -0700
- Subject: Recent tracepoint regressions
Hi,
While testing some patches this afternoon, I noticed that tracepoint
tests were causing gdb to segfault in several tests (actions.exp,
report.exp, tstatus.exp, tsv.exp, while-stepping.exp).
The following commit caused the regression:
commit ceff6a55e34396c81009e9c93086e7fa84eb73ec
Author: Tom Tromey <tromey@redhat.com>
Date: Mon Apr 22 16:46:14 2013 +0000
PR gdb/7912:
* Makefile.in (SFILES): Add filestuff.c
(COMMON_OBS): Add filestuff.o.
(filestuff.o): New target.
* auto-load.c (auto_load_objfile_script_1): Use
gdb_fopen_cloexec.
* auxv.c (procfs_xfer_auxv): Use gdb_open_cloexec.
* cli/cli-cmds.c (shell_escape): Call close_most_fds.
* cli/cli-dump.c (fopen_with_cleanup): Use gdb_fopen_cloexec.
* common/agent.c (gdb_connect_sync_socket): Use
gdb_socket_cloexec.
* common/filestuff.c: New file.
* common/filestuff.h: New file.
* common/linux-osdata.c (linux_common_core_of_thread)
(command_from_pid, commandline_from_pid, print_source_lines)
(linux_xfer_osdata_shm, linux_xfer_osdata_sem)
(linux_xfer_osdata_msg, linux_xfer_osdata_modules): Use
gdb_fopen_cloexec.
* common/linux-procfs.c (linux_proc_get_int)
(linux_proc_pid_has_state): Use gdb_fopen_cloexec.
* config.in, configure: Rebuild.
* configure.ac: Don't check for sys/socket.h. Check for
fdwalk, pipe2.
* corelow.c (core_open): Use gdb_open_cloexec.
* dwarf2read.c (write_psymtabs_to_index): Use gdb_fopen_cloexec.
* fork-child.c (fork_inferior): Call close_most_fds.
* gdb_bfd.c (gdb_bfd_open): Use gdb_open_cloexec.
* inf-child.c (inf_child_fileio_readlink): Use gdb_open_cloexec.
* linux-nat.c (linux_nat_thread_name, linux_proc_pending_signals):
Use gdb_fopen_cloexec.
(linux_proc_xfer_partial, linux_proc_xfer_spu): Use
gdb_open_cloexec.
(linux_async_pipe): Use gdb_pipe_cloexec.
* remote-fileio.c (remote_fileio_func_open): Use
gdb_open_cloexec.
* remote.c (remote_file_put, remote_file_get): Use
gdb_fopen_cloexec.
* ser-pipe.c (pipe_open): Use gdb_socketpair_cloexec,
close_most_fds.
* ser-tcp.c (net_open): Use gdb_socket_cloexec.
* ser-unix.c (hardwire_open): Use gdb_open_cloexec.
* solib.c (solib_find): Use gdb_open_cloexec.
* source.c (openp, find_and_open_source): Use gdb_open_cloexec.
* tracepoint.c (tfile_start): Use gdb_fopen_cloexec.
(tfile_open): Use gdb_open_cloexec.
* tui/tui-io.c (tui_initialize_io): Use gdb_pipe_cloexec.
* ui-file.c (gdb_fopen): Use gdb_fopen_cloexec.
* xml-support.c (xml_fetch_content_from_file): Use
gdb_fopen_cloexec.
* main.c (captured_main): Call notice_open_fds.
gdbserver
* Makefile.in (SFILES): Add filestuff.c.
(OBS): Add filestuff.o.
* Makefile.in (SFILES): Add filestuff.c
(COMMON_OBS): Add filestuff.o.
(filestuff.o): New target.
* auto-load.c (auto_load_objfile_script_1): Use
gdb_fopen_cloexec.
* auxv.c (procfs_xfer_auxv): Use gdb_open_cloexec.
* cli/cli-cmds.c (shell_escape): Call close_most_fds.
* cli/cli-dump.c (fopen_with_cleanup): Use gdb_fopen_cloexec.
* common/agent.c (gdb_connect_sync_socket): Use
gdb_socket_cloexec.
* common/filestuff.c: New file.
* common/filestuff.h: New file.
* common/linux-osdata.c (linux_common_core_of_thread)
(command_from_pid, commandline_from_pid, print_source_lines)
(linux_xfer_osdata_shm, linux_xfer_osdata_sem)
(linux_xfer_osdata_msg, linux_xfer_osdata_modules): Use
gdb_fopen_cloexec.
* common/linux-procfs.c (linux_proc_get_int)
(linux_proc_pid_has_state): Use gdb_fopen_cloexec.
* config.in, configure: Rebuild.
* configure.ac: Don't check for sys/socket.h. Check for
fdwalk, pipe2.
* corelow.c (core_open): Use gdb_open_cloexec.
* dwarf2read.c (write_psymtabs_to_index): Use gdb_fopen_cloexec.
* fork-child.c (fork_inferior): Call close_most_fds.
* gdb_bfd.c (gdb_bfd_open): Use gdb_open_cloexec.
All of these failures are occurring in trace_save, where it appears that
a call to start the writer was accidentally removed. The ChangeLog makes
no mention of this change, so I presume this was not intended.
Simply adding this back into it removes all these regressions.
diff --git a/gdb/tracepoint.c b/gdb/tracepoint.c
index f5c9c49..e2b21af 100644
--- a/gdb/tracepoint.c
+++ b/gdb/tracepoint.c
@@ -3341,6 +3341,8 @@ trace_save (const char *filename, struct
trace_file_writer *writer,
target is losing, we can get out without touching files. */
status = target_get_trace_status (ts);
+ writer->ops->start (writer, filename);
+
writer->ops->write_header (writer);
/* Write descriptive info. */
Keith