package install can print confusing error messages

[BrentBaccala]

For example, while installing the kash3 package, the output includes the following lines:

usage: sage-spkg-uninstall [-h] [-v] [-k] spkg [sage_local]
sage-spkg-uninstall: error: argument spkg: invalid spkg_type value: 'kash3'

This is due to the uninstall script throwing incorrect error types.
After applying this ticket's patch, the output reads:

usage: sage-spkg-uninstall [-h] [-v] [-k] spkg [sage_local]
sage-spkg-uninstall: error: argument spkg: 'kash3' is not a known spkg

Which is at least more sensible. kash3 is not known because it is an old-style package that didn't create a directory in build/pkgs.

CC: @embray @slel

Component: packages: optional

Keywords: spkg, uninstall

Author: Brent Baccala

Branch/Commit: f0afed6

Reviewer: Erik Bray

Issue created by migration from https://trac.sagemath.org/ticket/27670

[BrentBaccala]

comment:1

(Wrong branch - fixed below)

[BrentBaccala]

Commit: c50153a

[BrentBaccala]

Branch: public/27607

[BrentBaccala]

Author: Brent Baccala

[BrentBaccala]

Changed branch from public/27607 to public/27670

[BrentBaccala]

Changed commit from c50153a to f0afed6

[BrentBaccala]

Changed branch from public/27670 to none

[BrentBaccala]

Changed commit from f0afed6 to none

[BrentBaccala]

New commits:

f0afed6

Trac #27670: throw correct error types in uninstall script

[BrentBaccala]

Commit: f0afed6

[BrentBaccala]

Branch: public/27670

[slel]

Changed keywords from none to spkg, uninstall

[embray]

comment:6

This fix makes sense in its own right so +1.

Support for old-style SPKGs is deprecated and should be removed completely.

The only reason it hasn't been completely removed is last time the question came up there was paralysis about what to do about one particularly large package that would need to be hosted somewhere if it were converted to a new-style SPKG.

In that case it was just a large database, and is so large in fact that I'm not convinced downloading the whole database is the right thing to do in most cases anyways. It should be wrapped in a little queryable web service hosted online somewhere (perhaps even sagemath.org). Someone just needs to do that. I don't think that one package is reason enough to hold up removal of old-style SPKG support though.

[embray]

Reviewer: Erik Bray

[vbraun]

Changed branch from public/27670 to f0afed6