[poppler] [RFC] Replace GooHash by std::unordered_map

[Albert Astals Cid]

El dijous, 3 de maig de 2018, a les 0:57:21 CEST, Albert Astals Cid va 
escriure:
> El dimecres, 2 de maig de 2018, a les 19:11:19 CEST, Adam Reichold va
> 
> escriure:
> > Hello,
> > 
> > Am 01.05.2018 um 23:57 schrieb Albert Astals Cid:
> > > El diumenge, 4 de març de 2018, a les 22:59:20 CEST, Albert Astals Cid
> > > va
> > > 
> > > escriure:
> > >> El dissabte, 3 de març de 2018, a les 20:51:55 CET, Adam Reichold va
> > > 
> > > escriure:
> > >>> Hello Albert,
> > >>> 
> > >>> please note that in the previous patch, I made a mistake with the move
> > >>> constructor (the ref count should not be moved/swapped since it is
> > >>> tied
> > >>> to the instance, not its value). Attached is a fixed patch.
> > >>> 
> > >>> I found the issue during further performance tests which however only
> > >>> continued to strengthen to zero sum result. In any case, I extend the
> > >>> Python-based perftest script with a C++-based driver to be able to
> > >>> reliably measure single-page runtimes for this, which I will also
> > >>> attach
> > >>> if anyone is interested in this.
> > >> 
> > >> Ok, so it seems there's a mild agreement (or at least no strong
> > >> disagreement) that this may be the way forward. I will first wait to
> > >> get
> > >> poppler 0.63 out of the door (that is already late several months,
> > >> between
> > >> freedesktop shutting down ssh access, them forgetting to tell us it was
> > >> open again, my disk dying and then me procastinating as usual) and once
> > >> it's out I'll come back and reevaluate this.
> > > 
> > > I was trying to apply the patch for running a quick regtest but running
> > > git am over it gave me some warning/error I wasn't sure how to get out
> > > of
> > > correctly.
> 
> regtest results are good, i guess i'll have a second quick look at the
> patches and unless someone disagrees i'll push them.

It's in :)

Cheers,
  Albert