[darcs-users] [patch72] resolve issue1624 break global cache up into subdirectories

Trent W. Buck twb at cybersource.com.au
Mon Jan 11 06:41:20 UTC 2010


Isaac Dupree <ml at isaac.cedarswampstudios.org> writes:

> Trent W. Buck wrote:
>> Reinier Lamers <tux_rocker at reinier.de> writes:
>>> Op woensdag 23 december 2009 12:00 schreef je:
>>>> Have you already applied this? I may have found a bug on it.
>>> I considered applying it tonight, but it raised some more questions on IRC:
>>>
>>>   [21:26] <Heffalump> it effectively throws away the existing cache, right?
>>
>> My ~/ (including ~/.cache/darcs) is shared between hosts over NFS.
>> Suppose that host a has darcs 2.0.2, b has 2.3.1 and c has 2.4.0.  Does
>> this mean that
>>
>>     c$ darcs get http://darcs.net
>>
>> will delete all the cache files that a and b understand?  This would be
>> undesirable -- it'd make darcs on a and b slower.
>
> it's common practice -- annoying, but common -- for ~/.dot-files to
> change format as the program's version advances, (and often not be
> able to go back to a previous format), so that it's only practical to
> use one version of the program in any one stretch of time (and often
> only in monotonically increasing order of version).  For example,
> Firefox. And a lot of programs, at various version numbers.

I suspect that Firefox is an exception, not an exemplar.

I've run significantly different versions of the following systems
against the same set of dotfiles, without noticable problems:

    aptitude, bash[0], crawl, curl, darcs, devscripts, dillo, emacs[0],
    festival, git, gnupg, gtk, hg, lesspipe, mutt, nano[1], python,
    ratpoison, rcirc[0], readline, reportbug, screen[1], urxvt, vim[1],
    w3m and xterm

Systems that I can remember complaining include mg, offlineimap, xmonad,
xzgv and yi.

[0] These required dispatching on version number in places.

[1] I don't consider warning about being compiled without e.g. syntax
    highlighting, then carrying on anyway, to be significant.

> I think this is because it's usually infeasibly [...]

Maybe that's the case if you're a behemothic browser that treats Unix
users as second-class citizens...



More information about the darcs-users mailing list