This is the mail archive of the
mailing list for the Cygwin project.
Re: [PATCH setup 0/3] Setup replacement for incver_ifdep
- From: Achim Gratz <Stromeko at nexgo dot de>
- To: cygwin-apps at cygwin dot com
- Date: Wed, 23 Sep 2015 20:09:22 +0200
- Subject: Re: [PATCH setup 0/3] Setup replacement for incver_ifdep
- Authentication-results: sourceware.org; auth=none
- References: <1442937170-17580-1-git-send-email-jon dot turney at dronecode dot org dot uk> <87mvwegy0r dot fsf at Rainer dot invalid> <5602DEBA dot 9080601 at dronecode dot org dot uk>
Jon Turney writes:
> [replying to the right list, this time]
I was slightly bewildered by that former replyâ :-)
> On 22/09/2015 18:32, Achim Gratz wrote:
> I thought that permanent postinstall scripts run even when no packages
> are being installed, or only packages are being removed, so they they
> are effectively run every time setup is?
Yes. That's their virtue and limitation.
>>> Unfortunately, it's not very practical to change to doing that for the all
>>> packages which contain info files, so I am suggesting this approach.
>> So just do it unconditionally; I was planning to change _update-info-dir
>> accordingly, but haven't found the time yet. I do the same with mandb
>> locally already and I haven't looked back.
> Tempting, but I don't believe that is a good solution, since it adds
> the time it takes to rebuild the info directory to every setup run.
As done with autorebase it should be possible to track if there's
anything to do (maybe even using the same mechanism of looking in
/etc/setup). But unlike autorebase the potential benefit is limited to
those 18 or so seconds (give or take, depending on what kind of disk you
>> $ time /etc/postinstall/update-info-dir.sh.done
>> real 0m18.169s
>> user 0m3.261s
>> sys 0m5.703s
> I also don't think it sets a good example. Any other packages which
> require update scripts to run (which may take even longer) when
> packages are installed/removed, should not be doing that every time
> setup is run, if avoidable.
Yes, but again, that requires either a lot more smarts of setup.exe or a
lot more smarts from the package maintainer(s). Look at TeXLive for the
sort of things you'd need to do. As long as things are packaged via
cygport those smarts needed by a lot of packages could be implemented
there and info directory maintenance certainly belongs into that group.
But if you go back to the original specification when I proposed the
stratified postinstall, you'll see that I did provide for triggers and
some more things, but that didn't get implemented since the idea was
mooted at the time. We'd still need to have a bit of a discussion of
what exactly these should do and how they are supposed to be used before
an implementation can be attempted.
ANother idea I keep on the back-burner is to have a stratum that would
background the postinstall scripts that are running on it, so they
(optionally) don't keep setup.exe from exiting. You can only do that
with scripts that do non-essential tasks, but both infodir and mandb
housekeeping fall into that category.
+<[Q+ Matrix-12 WAVE#46+305 Neuron microQkb Andromeda XTk Blofeld]>+
Factory and User Sound Singles for Waldorf Q+, Q and microQ: