Remove sage-location's "sage-force-relocate" mechanism, fix script to work without SAGE_ROOT

[mkoeppe]

#21783 removes the last productive bit of sage-location. The only thing that sage-location does now is write a warning if it notices that the tree has moved.

We want to get rid of the whole mechanism completely.... (or reuse its remains to implement #31076).

In this ticket, we remove a part of it, the "sage-force-relocate" mechanism, which used to be invoked by sage-spkg.

We also fix the script so that it works in installations without SAGE_ROOT. This is for the benefit of downstream packaging and #30913.

(When we finally get rid of sage-location, we can finally close tickets #15146, #17479, #11755.)

This is also preparation of sorts for #31076, which proposes to add a different relocation mechanism.

Depends on #21783

CC: @jhpalmieri @culler @dimpase @kiwifb @antonio-rojas @mwageringel

Component: relocation

Author: Matthias Koeppe, John Palmieri

Branch/Commit: 3035932

Reviewer: John Palmieri, Matthias Koeppe

Issue created by migration from https://trac.sagemath.org/ticket/31270

[mkoeppe]

Branch: u/mkoeppe/remove-sage-location

[sagetrac-git]

Commit: 287dc70

[sagetrac-git]

Branch pushed to git repo; I updated commit sha1. New commits:

287dc70

src/doc/it/faq/faq-usage.rst: rimuovi gli antichi consigli relativi alla posizione della salvia

[mkoeppe]

Description changed:

--- 
+++ 
@@ -1,6 +1,6 @@
 #21783 removes the last productive bit of `sage-location`.
 
-We get rid of the whole mechanism completely.
+We get rid of the whole mechanism completely.... (or reuse its remains to implement #31076).
 
 When we get rid of `sage-location`, we can finally close tickets #15146, #17479, #11755)
 

[mkoeppe]

Author: Matthias Koeppe

[mkoeppe]

Description changed:

--- 
+++ 
@@ -1,6 +1,8 @@
-#21783 removes the last productive bit of `sage-location`.
+#21783 removes the last productive bit of `sage-location`. The only thing that `sage-location` does now is write a warning if it notices that the tree has moved.
 
-We get rid of the whole mechanism completely.... (or reuse its remains to implement #31076).
+We want to get rid of the whole mechanism completely.... (or reuse its remains to implement #31076). 
+
+In this ticket, we remove a part of it, the "sage-force-relocate" mechanism, which used to be invoked by `sage-spkg`.
 
 When we get rid of `sage-location`, we can finally close tickets #15146, #17479, #11755)
 

[sagetrac-git]

Changed commit from 287dc70 to b0e4816

[sagetrac-git]

Branch pushed to git repo; I updated commit sha1. New commits:

1dfc920	src/bin/sage-location: Do not fail if SAGE_ROOT is not set
b0e4816	src/bin/sage-location: Do not output 2 lines announcing a no-op

[mkoeppe]

Description changed:

--- 
+++ 
@@ -4,7 +4,9 @@
 
 In this ticket, we remove a part of it, the "sage-force-relocate" mechanism, which used to be invoked by `sage-spkg`.
 
-When we get rid of `sage-location`, we can finally close tickets #15146, #17479, #11755)
+We also fix the script so that it works in installations without `SAGE_ROOT`. This is for the benefit of downstream packaging and #30913.
 
-This is preparation of sorts for #31076, which proposes to add a different relocation mechanism.
+(When we finally get rid of `sage-location`, we can finally close tickets #15146, #17479, #11755.)
 
+This is also preparation of sorts for #31076, which proposes to add a different relocation mechanism.
+

[jhpalmieri]

comment:7

• I don't know how to test whether this will "fix script to work without SAGE_ROOT". Maybe the distro people have to look at that part.

• The script's main purpose is to see if Sage has been relocated and then to quit with an error message if so. Do we need to do this? Can we just document that it can't be moved and let it fail however if fails if someone ignores this? That is, can we delete this entirely?

[mkoeppe]

comment:8

Replying to @jhpalmieri:

• I don't know how to test whether this will "fix script to work without SAGE_ROOT". Maybe the distro people have to look at that part.

Yes, I hope this ticket can reduce the amount of downstream patching

[jhpalmieri]

comment:9

Also: would the check

    if OLD_SAGE_ROOT != SAGE_ROOT:

be more robust as

    if not os.path.samefile(OLD_SAGE_ROOT, SAGE_ROOT):

or SAGE_ROOT = os.pathlib.Path('...') followed by SAGE_ROOT.samefile(...) or something similar? Maybe we have to handle None as a special case.

Edit: typo: should be samefile, not samepath.

[mkoeppe]

comment:10

Replying to @jhpalmieri:

• The script's main purpose is to see if Sage has been relocated and then to quit with an error message if so. Do we need to do this? Can we just document that it can't be moved and let it fail however if fails if someone ignores this? That is, can we delete this entirely?

I think there's value in giving a clear error message - removing it could cause more traffic on sage-support

[mkoeppe]

comment:11

Replying to @jhpalmieri:

Also: would the check

    if OLD_SAGE_ROOT != SAGE_ROOT:

be more robust as

    if not os.path.samepath(OLD_SAGE_ROOT, SAGE_ROOT):

or SAGE_ROOT = os.pathlib.Path('...') followed by SAGE_ROOT.samefile(...) or something similar? Maybe we have to handle None as a special case.

Yes, this sounds like a good idea.

[mkoeppe]

comment:12

But in fact to make this code really useful, it should be revised to work with SAGE_LOCAL instead of SAGE_ROOT (see discussion in #30913). So on the present ticket, I wouldn't want to make such improvements