Front page | perl.perl5.porters |
Postings from January 2001
Alan Burlison <Alan.Burlison@uk.sun.com> writes: >Nick Ing-Simmons wrote: > >> Can those that fear breakage cook up tests that prove things >> work the way they expect (we could start with program from the >> 5.002 vintage archive Alan found). > >Note that Raphaels test script shows that the combination of eval and >signal handlers (the very reasom for using sigsetjmp) is broken anyway, >and as setjmp and sigsetjmp are not MT-safe they can't be used in >threaded perl. What do you think of noting the sigmask before we call a signal handler and restoring it if handler "dies"? (Leaving mask as handler leaves it if handler runs to bottom - so that handler changing mask is still valid.) That may not be quite what you would get with sigsetjmp() all the way down - but would seem to be a reasonable compromise. -- Nick Ing-SimmonsThread Previous | Thread Next