[PATCH] setup: implement "perpetual" postinstall scripts

Achim Gratz Stromeko@nexgo.de
Thu Feb 7 19:30:00 GMT 2013

Christopher Faylor writes:
> I don't really see how making setup.exe always run rebaseall makes
> unattended installs work better.

It requires less invocations of setup and yields a simpler install
script which consequently has less chance of being "repaired" to death
by someone ten timezones away when it finally gets rolled out.

> ???  If we need to rebase dlls then we should just run rebaseall.  There
> isn't any reason to tell people to run setup.exe to do that.

Right now you are telling users to run rebaseall after setup.exe in the
FAQ, actually.  Even with _autorebase, users really have to do that when
they have downgraded or re-installed packages and didn't remember to
also reinstall _autorebase, which they most likely didn't know about.
How to run rebaseall is described in another document that they then
have to find, comprehend and do.

Having setup.exe take care of that and additionally providing an easier
way to do the equivalent of rebase or rebaseall (the user is already
familiar with setup.exe at that point) does seem like an improvement to
me — if only because it requires less explanation and involves fewer
opportunities for operator error.

> If the problem is that people are installing third-party dlls then
> they should get in the habit of running rebaseall.  We don't want
> people superstitiously running setup.exe expecting it to solve their
> problems even when there are no new packages to install.

We are apparently talking about completely disjoint sets of users and I
can't hope for this discussion to lead any further.

> If we can't expect the unwashed to run autorebase then we could have
> an autorebase daemon running which just attempts to rebase dlls that
> it hasn't seen before every N minutes.

That idea could grow on me, but just like the proposal to somehow get
rid of setup.exe altogether it doesn't solve any problems in a shorter

> Sorry but I don't think this is a good idea.

Very well, apologies for bothering you then.

+<[Q+ Matrix-12 WAVE#46+305 Neuron microQkb Andromeda XTk Blofeld]>+

Samples for the Waldorf Blofeld:

More information about the Cygwin-apps mailing list