Re: Slowness of org-agenda-redo in org 9.7 vs 9.6.7

[dark.key8799@151e.ai]

On Fri, Jan 19, 2024, at 23:15, Ihor Radchenko wrote:
> I think that the missing lambdas should come from org.el.
> So, may you M-x eval-buffer org-agenda.el, org-element.el, and org.el.
>
> And do the same in Org 9.6. (there are also cryptic lambdas there in the
> profile)
>
> (In theory, you should not need to do this M-x eval-buffer trick if you
> run Org mode from "make repro" command line in the git repo. Not sure if
> it is easier for you or not)

Well bad news then good news.

Either by evaluating buffers as requested or by trying to use "make repro" (which I couldn't make work), I was still getting cryptic lambdas.

I ended reinstalling emacs from scratch and the slowness disappeared. So I guess some native-compiled files were not updated when I switched to Org 9.7 (even though I had nuked the straight folder). A profiler report of `org-agenda-redo' is also clean of lambdas (it does have #<compiled -...> statements but the call tree is decently readable).

Side note about "make repro": it wouldn't load org from the repo. Trying to understand why, running "make autoloads", I was expecting (perhaps wrongly) that `org-version.el' and `org-loaddefs.el' would be generated in the repo directory/sub-directories but that never happened (and make didn't output any error).
Launching "emacs -Q -l init.el" with init.el just having adding the repo to `load-path' and requiring org would fail on missing loaddefs.

Is there a standard procedure to 1/ follow when switching org version and/or 2/ clean all byte-compiled/native-compiled files in this kind of situation?

-- 
Alexandre Avanian