[Chandler-dev] --stderr and --nocatch

[Bryan Stearns]

Andi Vajda wrote:
> The purpose of --stderr is to output repository logger calls to 
> stderr. I don't think it's used for anything else.
I know I'm late to the party, but I'm pretty sure --stderr also prevents 
tracebacks from appearing in their own window in the UI, based on this 
comment in Chandler.py:
        # Redirect stdio and stderr to a dialog if we're running the debug
        # version. This is done to catch asserts, which otherwise will never
        # get seen by people who run Chandler using the launchers, e.g.
        # Aparna. If you're running release you can also set things up so
        # that you can see stderr and stdout if you run in a shell or from
        # wing with a console.
        #
        # useBestVisual - uses best screen resolutions on some old 
computers.
        #                 See wxApp.SetUseBestVisual

        redirect = not Globals.options.stderr
        app = wxApplication(redirect=redirect, useBestVisual=True)

Since we want tracebacks in the log, I think we still want --stderr when 
running automated tests (both on tbox and from do_tests).

Generally, I don't think too much info in the logs when running 
automated tests is a problem: I want to see as much info in the logs as 
I can get, because it may save me from wasting time trying to reproduce 
a problem. Tinderbox already uses a filter mechanism (which I've tweaked 
a couple of times) that pulls important information to the top of the 
log page, which makes it a lot easier to find specific failures -- if 
there's info about the failure in the log, that is.

> Getting rid of command line options just for the sake of getting rid 
> of them is helping nobody. End-users don't use command line options. 
> Some developers do, others can safely ignore them. 
+1

...Bryan