Synopsis - Issue Tracker

The function "exit()" does not belong to the list of async-signal-safe functions.
See section "2.4.3 Signal Actions" from the document "2.4 Signal Concepts".
http://opengroup.org/onlinepubs/009695399/functions/xsh_chap02_04.html#tag_02_04_03
http://opengroup.org/onlinepubs/009695399/functions/_Exit.html

I imagine that the output to "std::cerr" in a signal handling function may be also unsafe.

Should the corresponding calls be replaced by safe alternatives in your function "sighandler"?
http://synopsis.fresco.org/viewsvn/Synopsis/trunk/src/Support/ErrorHandler.cc?rev=1811&view=auto

Just curious..
So, instead of exit() would you suggest abort()ing?

Which "safe alternative" do you have in mind to use instead?

The I/O subsystem (usually libc) should not be affected by this.

Can the function "sighandler" be completely omitted?

Otherwise, consider to call "_Exit()".

Surely, if we wouldn't need the sighandlers, we would omit them.

Are you going to make your logging approach async-signal-safe?

So you just suggests to omit running the __exit_function_table.

How do you think about the message by Eric P. McCoy on the topic "threads with signals"?
http://groups.google.de/group/comp.os.linux.development.apps/msg/feb871fc978018f6

The object "std::cerr" must not be used inside a signal handler.

I, personally, think that is perfectly fine to use STDIO in a signal handler.

The worst that could happen is that there are a number of faults, which is fine since we're dead anyways.

The following instructions will work.

char const * message[] = "bye bye\n";
write(file, message, sizeof(message));

Better:
char const message[] = "bye bye\n";

Not sure how your c++ library implements string operator <<

Mine works fine, so perhaps you should rather fix your libc++ impl

Are you sure that the implementation of the output operator is async-signal-safe on all supported platforms?

Will it call "new" or "malloc()"?