Has bzip2 ever been both available and unusable?

[SnarkBoojum]

I'm surprised the bzip2 spkg-install doesn't start by checking if bzip2 is in the path... and just bail out if so: why build it anyway? As asked in the summary: was there ever a case where the system bzip2 wasn't suitable for sage?

Component: build

Reviewer: Jeroen Demeyer

Issue created by migration from https://trac.sagemath.org/ticket/14849

[nexttime]

Replying to @SnarkBoojum:

Was there ever a case where the system bzip2 wasn't suitable for sage?

I strongly doubt that... ;-)

As with many other spkgs, Sage simply doesn't bother to use a system-wide version.

Ceterum censeo Sage should (ship and) use xz. (And make bzip2 a prerequisite + an optional spkg.)

Btw., zlib and similar are (redundantly) included in many upstream packages...

[kiwifb]

comment:2

System bzip2 has probably always be suitable... provided that all the interesting bits are installed. I am guessing headers may be required and I am willing to bet that the library may not always installed by default (and that may have been more frequent in the past).

[SnarkBoojum]

comment:3

1. Do we really need the headers? [for python perhaps?]
2. If bzip2 is here, doesn't that mean libbzip2 is here too?

[kiwifb]

comment:4

I cannot talk about bzip2 in particular but I take care of some SLES servers in a professional capacity. I have seen libraries shipped as part of the dev package rather than the main package. The main executable shipped can very well be statically linked for speed.

[jdemeyer]

comment:5

Replying to @SnarkBoojum:

1. Do we really need the headers? [for python perhaps?]

I know for sure that Python checks for and links against the bzip2 library. And I guess that the non-Darwin implementation of sage --pkg uses this.

[jdemeyer]

comment:7

Can this be closed as "wontfix"?

[jdemeyer]

Reviewer: Jeroen Demeyer