bug#41646: Startup in Windows is very slow when load-path contains many

[Stefan Monnier via "Bug reports for GNU Emacs, the Swiss army knife of text editors" <bug-gnu-emacs@gnu.org>]

> [I have unarchived Bug#41646 so that this discussion is archived in the
> bug tracker.]

[ Note: I haven't seen the earlier part of this discussion (but I did read the
  old bug#41646 thread).  ]

>> Totally understand, and thanks for inviting Stefan and Andrea in the
>> conversation, to let new participants know the context quickly, I had
>> post a patch to speed up Emacs for the scenario that Emacs will very
>> slow on startup with hundreds packages (>300)  installed, the keypoint
>> is hundreds packages will add hundreds paths into the `load-path',
>> then a simple `(require X)' may trigger hundreds searching according
>> to the `load-path',

Yup, that's a known problem which we've been not solving for
a long time.
The "latest" workaround is `package-quickstart`, which reduces some of
the associated pain.

If you're not using `package-quickstart` and are bothered by a slow
startup, then I'd recommend you go back and enable `package-quickstart`.

Similarly in the old bug#41646 thread, I see mentions of 40s startup
time, 8s of which are *not* spent in openp: 8s is still quite slow, so
it might be worth looking at what the startup file is doing and see if
we could do less at startup (e.g. load things more lazily).

Of course, the "long load-path" problem will still end up biting in
some cases.  Until now we've managed to make those cases rare enough
that we haven't had to actually solve it.

>> so the patch build a map for feature --> filepath from the variable
>> `load-history' and store to disk, after that loading

I don't think we want a cache that's stored on disk: it would take too
much effort to create it, load it, ensure it's not damaged if several
Emacs sessions try to write it at the same time, make sure it's fresh,
etc...

Especially since I believe that building the cache shouldn't take very
long: longer than a single "look for file F in `load-path`", maybe, but
not by much.

IOW we could keep a cache that's populated on-the-fly the first time we
`load` something, and that's then automatically refreshed when `load`
sees a new `load-path`.


        Stefan