PR #3286 merged into devel. Please continue testing

[Edward K. Ream]

PR #3286 is now in devel. It restores some legacy functions and fixes a missing alias.

Please continue testing. The latest code fixes several brain spikes on my part, but more problems may remain.

Edward

[Edward K. Ream]

On Tuesday, April 18, 2023 at 10:46:39 AM UTC-5 Edward K. Ream wrote:

PR #3286 is now in devel. It restores some legacy functions and fixes a missing alias.

Please continue testing. The latest code fixes several brain spikes on my part, but more problems may remain.

Meh.  g.os_path_expanduser does not (now) finalize its result. That may be wrong, but I can't be sure yet.

Edward

[Thomas Passin]

This has fixed the g.os_path_join() problem.

[Edward K. Ream]

On Tue, Apr 18, 2023 at 11:02 AM Thomas Passin <tbp1...@gmail.com> wrote:

This has fixed the g.os_path_join() problem.

Whew.  Thanks for the testing.

I've made copies of the g.os_path_expanduser and g.os_path_join as they existed in Leo 6.7.2.  We can't go back to the exact legacy code, but I'll take yet another look after I take a break.

Edward

[Edward K. Ream]

On Tuesday, April 18, 2023 at 11:10:29 AM UTC-5 Edward K. Ream wrote:

I've made copies of the g.os_path_expanduser and g.os_path_join as they existed in Leo 6.7.2.  We can't go back to the exact legacy code, but I'll take yet another look after I take a break.

As of rev ca92d1f in devel, g.os_path_expanduser calls os.path.normpath, as in Leo 6.7.2.

This change should ensure that all legacy scripts and plugins work exactly as before.

As always, please report any problems.

Edward

P.S. g.os_path_join needs no further changes. It should be exactly equivalent to the legacy version.

EKR

[Thomas Passin]

g.os_path_join() is still doing the right thing on my system.

[Thomas Passin]

This whole exercise puts me in mind of the several levels of testing: unit, integration, and system.  Unit tests can only take you so far.  In the case of these os.path-like functions, a unit test could prove that the output equals the intended output, but it could not show that all consuming modules expect the function to act that way.

[Edward K. Ream]

On Tue, Apr 18, 2023 at 12:03 PM Thomas Passin <tbp1...@gmail.com> wrote:

This whole exercise puts me in mind of the several levels of testing: unit, integration, and system.  Unit tests can only take you so far.  In the case of these os.path-like functions, a unit test could prove that the output equals the intended output, but it could not show that all consuming modules expect the function to act that way.

Yes. And this exercise shows the limits of waiting for approving reviews of PRs, hehe.

I think our travails were worthwhile. The new code, and the new names, are better.

Edward