On Wed, Apr 09, 2014 at 11:26:08AM -0700, Dave Huseby wrote:
> 3. It's painfully obvious that we have multiple solutions for the same
> problems.  It's also painfully obvious that it isn't easy for developers
> to find out about existing tools (I know, I've been digging for days)
> and therefore build their own solution under the false pretense that a
> tool doesn't already exist.  (I'm guilty of this myself, see #2).

There's also some unfortunate duplication with respect to sending
information in the opposite direction.

For example, the B2G about:memory support claims three realtime signals
(one each for "dump report", "dump report after minimizing usage",
and "dump GC/CC logs"), which are sent by a special utility we build
because the existing kill(1) can't deal with numbers >= 32, and then the
pipe-to-self trick is used to get out of signal handler context, etc.
Followed by the previously mentioned polling for files to appear, which
results in hacks like MOZ_IGNORE_NUWA_PROCESS that annoy people and
reduce their productivity.

Except we already have Marionette and XPConnect, which (now that bug
973090 landed) can replace all of that.  I think.  The only inelegance
is that I'd like JS callbacks to call up into Python on the host, but
execute_script can only return a value; this is workaroundable, but it'd
be nice if there already were some way to yield a stream of events....

The situation with the profiler is similar, except the reinvented wheel
is IMHO a bit more angular.  There might be some interest in removing
the overt async signal unsafety, the custom sort-of-CSV parser, etc.

--Jed

_______________________________________________
dev-b2g mailing list
[email protected]
https://lists.mozilla.org/listinfo/dev-b2g

Reply via email to