Allow sage -i python to actually install the right version of Python

[jdemeyer]

CC: @jhpalmieri

Component: scripts

Branch/Commit: u/jdemeyer/ticket/19105 @ 74c78c1

Reviewer: Jeroen Demeyer

Issue created by migration from https://trac.sagemath.org/ticket/19105

[vbraun]

comment:1

I don't understand what the ticket proposes.

• Renaming python2 -> python is backward
• Renaming python3 -> python is too soon
• Adding more spaghetti code to the install script to special-case python sucks
  IMHO the current behavior is ideal, throw an error and point the user to either python2 or python3

[jdemeyer]

comment:2

Replying to @vbraun:

I don't understand what the ticket proposes.

• Renaming python2 -> python is backward
• Renaming python3 -> python is too soon
• Adding more spaghetti code to the install script to special-case python sucks

Luckily, none of the above applies.

There is already code in build/make/install which deals with Python 2 vs. Python 3. The resulting file build/make/Makefile contains what is needed to make this work very easily.

[jdemeyer]

comment:3

Admittedly, making ./sage -f python work is a lot harder than making ./sage -i python work. Unfortunately, it is ./sage -f python which is actually needed (for example, after installing openssl).

[vbraun]

comment:4

IMHO it is not needed, just make up your mind which python version you want to install.

[jdemeyer]

Changed dependencies from #19101, #19103, #19104 to #19119

[jdemeyer]

Branch: u/jdemeyer/ticket/19105

[jdemeyer]

New commits:

93ec09b	Change sage -p to always install a package
c6ddd0e	Better help
f3d7ef3	Remove -f option to sage-spkg when running sage -i
752e269	Use PKG-clean target to implement sage -f PKG
1812e4f	Merge tag '6.9.beta6' into t/19119/ticket/19119
74c78c1	Allow "sage -i python" and "sage -f python"

[jdemeyer]

Commit: 74c78c1

[jdemeyer]

comment:8

Note that this is the only commit for this ticket, the rest comes from dependencies:

New commits:

74c78c1

Allow "sage -i python" and "sage -f python"

[vbraun]

comment:9

So the solution is C: Adding more spaghetti code to the install script to special-case python

[jhpalmieri]

comment:10

Another possible solution: don't search on-line standard packages.

diff --git a/build/bin/sage-spkg b/build/bin/sage-spkg
index f4b085b..356efb9 100755
--- a/build/bin/sage-spkg
+++ b/build/bin/sage-spkg
@@ -331,7 +331,7 @@ if [ ! -f "$PKG_SRC" ]; then
             if [ $? -ne 0 ]; then
                 exit 1
             fi
-            for repo in optional experimental standard huge archive; do
+            for repo in optional experimental huge archive; do
                 # Download the list of packages.
                 echo ">>> Checking online list of $repo packages."
                 # File inside DOT_SAGE should be writable

That's where the problematic Python package is, right? And with recent versions of Sage, is there any reason to search "standard" packages? I think that all of those are either outdated or have more recent new-style replacements.

[jdemeyer]

comment:11

Replying to @jhpalmieri:

And with recent versions of Sage, is there any reason to search "standard" packages?

As you recall from #19004, we aren't searching anything anymore.

There are two partially separate issues here:

1. make sure that sage -i python does not install the wrong Python package.
2. make sure that sage -i python does install the correct Python package.

This patch here is more about the second point.

[jhpalmieri]

comment:12

There was also talk of rolling back #19004, in which case my suggestion would apply. But I guess not here. Is that a worthwhile change to make in any case? Should I open a ticket?

[jdemeyer]

comment:13

Replying to @jhpalmieri:

Should I open a ticket?

For what exactly?

[jhpalmieri]

comment:14

Never mind. If you take any efforts to roll back #19004, consider making those changes.

[jdemeyer]

comment:15

Replying to @jhpalmieri:

If you take any efforts to roll back #19004

Well, there is this huge thread on sage-devel discussing that. Some people are in favour, some against, no clear consensus. There is also a ticket for the rollback: #19158.

[jdemeyer]

Changed dependencies from #19119 to none

[jdemeyer]

Reviewer: Jeroen Demeyer

[jdemeyer]

Changed author from Jeroen Demeyer to none

[jdemeyer]

comment:16

Close as "nobodycares".