Current state of trace mode?

[Joshua Root]

On 2012-9-1 08:23 , Clemens Lang wrote:
> Hey,
> 
> does anybody know what the current state of trace mode in MacPorts is?

Pretty sure it's at least partially broken. The tests don't pass.
Haven't had the time to investigate properly.

> The code seems old, but contrary to my expectations the bit rod doesn't
> seem to have broken it completely (e.g., segfaulting or similar).

Yeah, I fixed it a while back to cope with 64-bit inodes and the related
new syscalls.

>>From my tests, it seems most things work, but checks against
> dependencies fail, because the registry might not be open.

Interesting, I hadn't considered the registry changes might be a factor.

> Is anybody still using trace mode? Does anybody remember what
> performance impact trace mode had?

Probably nobody uses it because it doesn't work. It slowed things down
noticeably but not horribly.

> Is having a working trace mode a good idea? Should we try to get it back
> working and optimize it to a point where it can be turned on globally?
> (I'm thinking about avoiding conflicts with /usr/local, especially brew
> here.) Maybe it could also be used as replacement for "How to test" in
> http://trac.macports.org/wiki/UsingTheRightCompiler?

Yes, having a working and optimised trace mode would be great.

There was another thread discussing it at the end of last year, complete
with jkh threatening to write some code as I recall. But I guess nobody
found the time yet.

I'd be very interested to see if we can use the sandboxing tech in newer
OS versions as well as or instead of using darwintracelib. Last I
checked though there didn't appear to be a way to set up the
entitlements programmatically the way we would need.

- Josh