After changing catching signals in bash - added CATCH_SIGNALS () macro -
we aren't catching them after calling execve(). This short patch fix it:
diff -up bash-3.2/execute_cmd.c.execve_catch_signal bash-3.2/execute_cmd.c
--- bash-3.2/execute_cmd.c.execve_catch_signal 2008-09-15
12:20:18.000000000 +0200
+++ bash-3.2/execute_cmd.c 2008-09-15 12:20:55.000000000 +0200
@@ -3943,6 +3943,7 @@ shell_execve (command, args, env)
SETOSTYPE (0); /* Some systems use for USG/POSIX
semantics */
execve (command, args, env);
i = errno; /* error from execve() */
+ CATCH_SIGNALS ();
SETOSTYPE (1);
I'm sorry. I didn't explain what is the problem.
The problem is, when bash tries to run any process by calling execve()
from slow (maybe not only slow) hardware (for example mounted NFS).
Driver of this hardware says kernel, that it needs to wait and kernel is
waiting until signal handling is done. Probably I'm not saying it clearly.
+++ bash-3.2/execute_cmd.c 2008-09-16 14:52:14.000000000 +0200
@@ -99,6 +99,8 @@ extern int errno;
# include "bashhist.h"
#endif
+#include "quit.h"
+
extern int posixly_correct;
extern int breaking, continuing, loop_level;
extern int expand_aliases;
@@ -3943,6 +3945,7 @@ shell_execve (command, args, env)
SETOSTYPE (0); /* Some systems use for USG/POSIX
semantics */
execve (command, args, env);
i = errno; /* error from execve() */
+ CHECK_TERMSIG;
SETOSTYPE (1);
/* If we get to this point, then start checking out the file.
RR