GitLab

Use '--timeout' for attach/detach/eval.

See tmm1/rbtrace#32

Fix `rbtrace --exec`

We set RUBYOPT using a path relative to the current file when execing a child
process. The file that contains this logic was moved in
d502dde9 but the relative path wasn't updated
at that time.

Make sure that a tracer will exit when the socket has no data AND the tracee is gone

rbtrace works on rubies beyond 1.9

Update rbtrace.c

Pass libdir on to ext configure script

Use RBASIC_CLASS instead of RBASIC()->klass

Ensure that the path to the socket is cleaned up when detaching

Currently the path to the socket was only cleaned up on exit

Use libdir from RbConfig to detect lib vs lib64

Added definitions for PLATFORM (PLATFORM_FREEBSD, PLATFORM_LINUX) on lines 44-50.

Added setpgid(0,0); command for FreeBSD which should also be fine using Linux (except CentOS I am not sure) on lines 967 to 972. In this case I left the old command setpgrp(); for linux just in case. But in theory using 'setpgid(0,0);' should work on all platforms according to all documentation I've run into.

Anyway all the documentation for FreeBSD and Linux & other say:

 The call setpgrp() is equivalent to setpgid(0,0).

Please do check and Test. :-)

Fix `eval`

After the first time, @eval_result is already set, so the output
is the same/it doesnt wait for a response

Libify

Fix for issue 14: rbtrace ignores the value for --slow

Bump to 0.4.1