For strace to be really useful when run as part of an automated
testing system, it needs to behave like gdb does with signals, and
pass them on to its child process wherever possible - so it can be
inserted into a command in-place without disrupting the surrounding
code. Otherwise, it is really difficult to send signals to the child
process.

This behaviour should probably be optional, and not the default.

Hi,

I'd like to second this. I'm trying to find a bug in a program's SIGINT
handling, and strace's behaviour makes it pretty difficult for me to test.

   Simon

Hello,

a long while ago, you said that it would be nice if strace could pass
on the signals that it gets to its traced child to ease the delivery
of signals to the traced processes.

I think that in between (?), strace implemented the -D option which
is, according to the manpage:

  Run tracer process as a detached grandchild, not as parent of the
  tracee.  This reduces the visible effect of strace by keeping the
  tracee a direct child of the calling process.

In practice, the traced process will be the child of the launcher
process, as usual. So, I think that this option is sufficient to close
that bug report, don't you think?

Bye, Mt.