Re: [SECURITY] Async-signal-unsafe operations in alrm_catcher() signal handler

On 8/27/26 8:41 PM, Paul Eggert wrote:
> On 2026-08-27 16:22, [email protected] wrote:
>>     printf (_("\007timed out waiting for input: auto-logout\n"));
>>     fflush (stdout);
> 
> Thanks for the bug report. I see that the current Git version has evolved a 
> bit, in that it now calls 'write' (which is async-signal-safe) rather than 
> printf and fflush. However, it still calls gettext (which is not async- 
> signal-safe) and it calls jump_to_top_level (which is even less so), so 
> there's still a couple of bugs here.

The right thing to do is to restrict SIGALRM to cases when readline is not
being used, since readline can handle this directly without using signals,
and use a flag-based approach similar to how the shell handles other fatal
signals.

This code has survived virtually unchanged for 35 years; probably time to
bring it up to date.

Chet
-- 
``The lyf so short, the craft so long to lerne.'' - Chaucer
		 ``Ars longa, vita brevis'' - Hippocrates
Chet Ramey, UTech, CWRU    [email protected]    http://tiswww.cwru.edu/~chet/

-----BEGIN PGP SIGNATURE-----

wmMEABEIACMWIQR8ATX7CIqvbGbGULm7WGnwZOp0qwUCapHuJwUDAAAAAAAKCRC7WGnwZOp0qxY+
AJ0VYYX+YyE8D10SLk3BZY0314IDtACfaD7jlIdwv/SOamRn3k8oEuJ8DTM=
=dN/F
-----END PGP SIGNATURE-----