[Toybox] [PATCH] crond: Fix timing

[Rob Landley]

On 1/12/25 20:09, Kana Steimle wrote:
> The changes are working fine for me so far. But after trying more
> things with crond, I found a few more issues I wanted to fix, all
> related to environment variables. Another patch is attached.
> - When setting a variable in the crontab file, the value would have a
> newline at the end of it. That's because getline() returns the line
> with a newline at the end, and the code didn't strip that newline out.
> - The HOME variable was being sourced from what HOME was set to when
> crond was run, rather than what the user's home is in the passwd file.
> I believe the intention was to set HOME to what it is in the passwd
> file, but allow the crontab to override it.
> - I decided to also set LOGNAME to the username of the crontab's
> owner, since the posix manpage for crontab mentions that as one of the
> variables that should be set.

Applied, but:

1) the vfork() child shares the parent's mappings, so modifications to 
the memory affect the parent process when done from the child. I believe 
this includes the environment space, and thus changes to the environment 
variables from the child are visible/persistent in the parent.

2) setenv() mallocs a new string and leaks the old one, if any, so doing 
that a lot in a long-lived server process gradually expands its memory 
footprint until reboot. (This is why lib/env.c exists.)

3) I _think_ getpwnam() returns the same static structure over and over, 
but am not sure? (The man page isn't clear.)

But now that a user has showed up, I'll bump crond.c to the top of my 
cleanup list to promote for the release after the one I'm working on now...

Thanks,

Rob