OpenE2K/binutils-gdb
12
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity

Fix bug with gdb resending SIGSTOP to an attached process that was forcibly

Browse Source 
stopped during attach, thus stopping it again.
This commit is contained in:
Fred Fish 1992-04-03 19:29:29 +00:00
parent 4d0d3491e3
commit d65eee73d4
2 changed files with 27 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

8
gdb/ChangeLog
View File

@ -1,3 +1,11 @@
Fri Apr 3 11:23:03 1992 Fred Fish (fnf@cygnus.com)
* procfs.c (procinfo struct): Add nopass_next_sigstop member.
* procfs.c (attach): Set nopass_next_sigstop if attached
process is forcibly stopped.
* procfs.c (child_resume): Use nopass_next_sigstop to suppress
resending SIGSTOP to attached process on first resume.
Fri Apr 3 01:37:26 1992 Stu Grossman (grossman at cygnus.com)
* Makefile.in (SFILES_MAINDIR): add mipsread.c

20
gdb/procfs.c
View File

@ -77,6 +77,7 @@ struct procinfo {
int fd; /* File descriptor for /proc entry */
char *pathname; /* Pathname to /proc entry */
int was_stopped; /* Nonzero if was stopped prior to attach */
int nopass_next_sigstop; /* Don't pass a sigstop on next resume */
prrun_t prrun; /* Control state when it is run */
prstatus_t prstatus; /* Current process status info */
gregset_t gregset; /* General register set */
@ -2012,6 +2013,11 @@ attach (pid)
close_proc_file (&pi);
error ("PIOCSTOP failed");
}
pi.nopass_next_sigstop = 1;
}
else
{
printf ("Ok, gdb will wait for process %u to stop.\n", pid);
}
}
@ -2386,6 +2392,17 @@ NOTE
to be current. One case where this might be necessary is if the
user explicitly changes the PC value that gdb considers to be
current. FIXME: Investigate if this is necessary or not.
When attaching to a child process, if we forced it to stop with
a PIOCSTOP, then we will have set the nopass_next_sigstop flag.
Upon resuming the first time after such a stop, we explicitly
inhibit sending it another SIGSTOP, which would be the normal
result of default signal handling. One potential drawback to
this is that we will also ignore any attempt to by the user
to explicitly continue after the attach with a SIGSTOP. Ultimately
this problem should be dealt with by making the routines that
deal with the inferior a little smarter, and possibly even allow
an inferior to continue running at the same time as gdb. (FIXME?)
*/
void
@ -2396,7 +2413,7 @@ child_resume (step, signo)
errno = 0;
pi.prrun.pr_flags = PRSVADDR | PRSTRACE | PRSFAULT | PRCFAULT;
pi.prrun.pr_vaddr = (caddr_t) *(int *) &registers[REGISTER_BYTE (PC_REGNUM)];
if (signo)
if (signo && !(signo == SIGSTOP && pi.nopass_next_sigstop))
{
set_proc_siginfo (&pi, signo);
}
@ -2404,6 +2421,7 @@ child_resume (step, signo)
{
pi.prrun.pr_flags |= PRCSIG;
}
pi.nopass_next_sigstop = 0;
if (step)
{
pi.prrun.pr_flags |= PRSTEP;