| Message ID | CAGyQ6gwhzpzy6bVtMaj7NOqtXnKQDzOG2OdqesJUkya4bpAF9Q@mail.gmail.com |
|---|---|
| State | New |
| Headers | show |
On Sat, Nov 29, 2014 at 1:27 PM, Ulrich Weigand <uweigand@de.ibm.com> wrote: >> 2014-11-29 Siva Chandra Reddy <sivachandra@google.com> >> >> * eval.c (evaluate_subexp): Check that thread stack temporaries >> are not already enabled before enabling them. > > This is OK. Good catch! Thank you. I have now pushed it as 18ac6ffc9b119c8d55a1cd0339d47998076faf2b.
diff --git a/gdb/eval.c b/gdb/eval.c index a13793c..c2ab879 100644 --- a/gdb/eval.c +++ b/gdb/eval.c @@ -69,7 +69,8 @@ evaluate_subexp (struct type *expect_type, struct expression *exp, int cleanup_temps = 0; if (*pos == 0 && target_has_execution - && exp->language_defn->la_language == language_cplus) + && exp->language_defn->la_language == language_cplus + && !thread_stack_temporaries_enabled_p (inferior_ptid)) { cleanups = enable_thread_stack_temporaries (inferior_ptid); cleanup_temps = 1;
The regression was pointed out here: https://sourceware.org/ml/gdb-patches/2014-11/msg00728.html It was introduced by my patch which was pushed as 6c659fc2c7cd2da6d2b9a3d7c38597ad3821832a. What my patch did was to make evaluate_subexp non re-entrant. As in, if evaluate_subexp is called with *pos == 0 while a top level expression is under evaluation, then it will crash with symptoms similar to those reported above. The attached patch fixes this problem and hence the regression. Jan Kratochvil has kindly tested it for me and confirmed that it indeed fixes the regression and does not introduce any new ones. gdb/ChangeLog: 2014-11-29 Siva Chandra Reddy <sivachandra@google.com> * eval.c (evaluate_subexp): Check that thread stack temporaries are not already enabled before enabling them.