[PATCH 1 of 1] subrepo: check for dirty warns on missing revision instead of abort

[Matt Mackall]

On Fri, 2011-05-13 at 17:12 +0200, Friedrich Kastner-Masilko wrote:
> Zitat von Matt Mackall <mpm at selenic.com>:
> 
> > Sorry, .node() here needs to be .hex().
> 
> Changed that here, thanks for the update. It worked with .node(), too.

Well, no, it didn't. The test always failed because it tried to compare
a hex value to a binary and thus always claimed the subrepo was dirty.

> The previously written statement still applies, though. No warning on  
> "hg update" to a changeset containing the same subrepo hash. I guess  
> this is because the .hgsubstate is not in the applyaction-list and  
> thus the subrepo-system is not touched besides evaluating dirty() .

Are we talking about something else now?

The bug you originally reported was: hg aborts while checking dirty() if
the subrepo revision it's checking for has vanished. I believe my
current patch fixes that while leaving everything else unchanged, please
confirm.

If there are other issues remaining in this function, please describe
them separately. We are simple folk here and like to fix one thing at a
time.

-- 
Mathematics is the supreme nostalgia of our time.