[PATCH 1/3] stop treating waitpid syscall specially
roland at redhat.com
Wed Feb 11 02:33:18 UTC 2009
TBH, I am not really worried about oddities with Linux PID namespaces.
Tough titties for them that uses it, really. (But, that just means, worry
about it later and not too hard rather than call it a priority.) That
said, yes that wait magic is obsolete. Current kernels don't have the
problem. Possibly even no 2.4 kernel had the problem, though I'm not at
all sure about anything except fairly recent vintages of 2.6 (I know
much earlier 2.6 had code for this, but I don't know about its bugginess).
But, I really strongly tend towards erring on the conservative side with
everything in strace. That is, if it ain't (very, or newly) broke, don't
fix it. The wait magic is kind of broken, but not at all newly.
So I say leave it alone until someone really screams.
If we are going to get into things that can be simpler if we assume a
recent enough Linux kernel, there is a ton of cruft.
More information about the Strace-devel