Vladimir Kotal <Vladimir.Kotal at Sun.COM> writes: > Richard Lowe wrote: > > <snip> > >> The basic theory "Cache things we're explicitly told, and run faster" >> holds, though with somewhat associated risk. The precise method >> outlined above falls flat. It is imperative the active list as seen >> by other things (think reci) is *correct*. Letting people influence >> that correctness is only going to cause more, and worse, problems for >> all concerned. > > I did not propose to change the internal Mercurial active list, only > the active list viewed by Cadmium (and only for certain cdm > subcommands).
Mercurial doesn't have the concept of an active list, the 'hg list' output is all cdm. We use that list in recommit, it *must* be accurate. > Say webrev is left intact. Then the worst error which could happen > would be manual and real active list getting out of sync which would > be detected by running 'hg outgoing', 'hg pbchk'.. and in the produced > webrev. The goal of the proposed change is to speed up common > code-still-in-flight tasks. Right, as I've said in a prior incarnation of this thread, I think I have a workable scheme to deal with that, it's just not going to get done immediately. What is best considered a prototype of it does help, and beyond some particularly thorny implementation details, seems to function. I have hopes those issues can be resolved. > Or is there something else I am missing ? See above. -- Rich
