src/usr.bin/mail, branch main

mail: tests: reset signal disposition before testing

2025-05-29T20:08:50Z

As a good citizen of the OS, mail(1) won't try to catch signals that were ignored when it started and it won't reconfigure its signal mask unless it's going to try and handle a signal. The test should start mail(1) off in a well-known state for the signal that it's trying to test in order to get the behavior that's desired. No functional change, just improves the resilience of the test. Reviewed by: des Sponsored by: Klara, Inc. Differential Revision: https://reviews.freebsd.org/D50599

mail: don't rewrite buffer sizes as much

2025-05-29T20:08:49Z

Thes buffers are arrays with a known size, just use sizeof() rather than rewriting their sizes -- no functional change, slightly more resilient against future possible errors. Reviewed by: markj Sponsored by: Klara, Inc. Differential Revision: https://reviews.freebsd.org/D50582

mail: Add test cases for SIGHUP.

2025-05-14T19:31:47Z

MFC after: 1 week Sponsored by: Klara, Inc. Reviewed by: kevans Differential Revision: https://reviews.freebsd.org/D50296

mail: exit with the correct exit status on SIGHUP in send mode

2025-05-14T19:20:35Z

Motivated by POSIX conformance requirements: mailx(1) is expected to take the default action for every signal except SIGINT in interactive mode. We still handle other signals that we shouldn't based on the spec (e.g., SIGTSTP), but conforming there is not as straightforward as we do more than just cleanup in response. Note that when the spec says that we must take the default action, it does not mean so strictly. Namely, we *can* do some sensible cleanup if we'd like, but we enter into nonconformant territory if we don't reflect the termination via signal in our exit status. That is why this change doesn't actually remove the handler; we're still conformant as long as the end result is the same as if we took the default action. Reviewed by: des Sponsored by: Klara, Inc. Differential Revision: https://reviews.freebsd.org/D50251

mail: remove the SIGHUP handler from lex.c

2025-05-14T19:20:34Z

If the command is simply going to exit without doing any cleanup, then an immediate exit(3) does not add any value and in-fact obscures that the program was terminated by a signal. This is motivated by POSIX conformance requirements, but it's also a little bit of a tidy-up. Some investigation into the history of mail(1) shows that, at one point, hangup() would attempt to save messages into mbox before exiting. It was later transformed into a simple exit() and likely could have been GC'd at that point. Reviewed by: des, emaste, ivy Sponsored by: Klara, Inc. Differential Revision: https://reviews.freebsd.org/D50247

mail: Further refine the SIGINT test.

2025-05-06T07:49:45Z

* Wait at least 1-2 s before sending the second SIGINT. * If the child is still running after 15 s, send a SIGKILL. * Improve the exit status checks. Fixes: 59597032c948 MFC after: 1 week Sponsored by: Klara, Inc. Reviewed by: markj Differential Revision: https://reviews.freebsd.org/D50092

mail: Exit non-zero on failure to collect mail.

2025-04-29T15:12:30Z

While here, avoid non-portable `pipe2()` in the regression test. Fixes: 59597032c948 MFC after: 1 week Sponsored by: Klara, Inc. Reviewed by: kevans Differential Revision: https://reviews.freebsd.org/D50069

mail: Don't trap signals we shouldn't.

2025-04-27T06:29:32Z

When in interactive mode, trap SIGINT, SIGHUP, and tty-related signals. Otherwise, leave signals untouched as required by POSIX. MFC after: 1 week Sponsored by: Klara, Inc. Reviewed by: kevans Differential Revision: https://reviews.freebsd.org/D50011

Remove residual blank line at start of Makefile

2024-07-15T22:43:39Z

This is a residual of the $FreeBSD$ removal. MFC After: 3 days (though I'll just run the command on the branches) Sponsored by: Netflix

mail: add volatile in grabh()

2024-01-05T02:40:53Z

setjmp() requires that any stack variables modified between the setjmp call and the longjmp() must be volatile. This means that 'saveint' in grabh() must be volatile, since it's modified after the setjmp(). Otherwise, the signal handler is not properly restored, resulting in a crash (SIGBUS) if ^C is typed twice while composing. PR: 276119 Reported by: Christopher Davidson MFC after: 2 weeks Pull Request: https://github.com/freebsd/freebsd-src/pull/993