bug#63710: use-package not working

[Andrea Corallo]

Andrea Corallo <akrl@sdf.org> writes:

> Andrea Corallo <akrl@sdf.org> writes:
>
>> Pedro Andres Aranda Gutierrez <paaguti@gmail.com> writes:
>>
>>> First: the .emacs.d/init.el test file
>>>
>>> ------
>>> ;;
>>> ;; My .emacs file
>>> ;;
>>> (require 'package)
>>> (setq package-archives '(("melpa" . "https://melpa.org/packages/";)
>>> ("gnu" . "https://elpa.gnu.org/packages/";)
>>>                          ("melpa-stable" . 
>>> "https://stable.melpa.org/packages/";)
>>> ("nongnu" . "https://elpa.nongnu.org/nongnu/";)))
>>> (eval-when-compile
>>>   (require 'use-package))
>>> (unless (file-directory-p package-user-dir)
>>>     (package-refresh-contents))
>>> (require 'use-package-ensure)
>>> (setq use-package-always-ensure t)
>>>
>>> (use-package "compat")
>>> ------
>>>
>>> Second: 1d5b164109b59559d34c545c2a163fa067ca22b2 does work
>>>
>>> BTW,  it has to do with native compilation. I have accidentally not deleted 
>>> the eln-cache in on of the tests and the
>>> 'wrong' Emacs has started correctly:
>>> 1. started 1d5b164109b59559d34c545c2a163fa067ca22b2 with eln-clean cache 
>>> ---> OK
>>> 2. started b2b2be98da5825e6bc130999ffe38f5ed615586a with eln-cache from the 
>>> last start --> OK
>>> 3. started b2b2be98da5825e6bc130999ffe38f5ed615586a with clean eln-cache 
>>> ---> KO
>>
>>>From these SHA1s looks like there's good chance it's my fault...
>>
>> I'll have a look.
>
> Okay I did some investigation and understood the issue.
>
> `package-load-descriptor' was misscompiled and returning always nil as
> the compiler was not fully picking up the user defined types at compile
> time, it worked only after the first faulty compilation (thing that made
> the investigaiton a bit slower).
>
> I'm working on a more complete fix but this requires careful testing, I
> pushed now c8e527cbeee in order to have master stable again while I keep
> working on the most comprehensive fix.

Okay RE-sending from this account and ditching my sdf.org probably
forever as I'm getting a bunch of bounces classifying it again as spam:

Hi all,

I pushed what I think is the final fix and reverted the temporary one.

As of fbc96dcc4ea I'm not aware of any issue with package.el or other
code.

Please let us know if works for you and we can close the bug or, if any
other issue arise (please remember to clean the eln-cache before testing
as it may contain misscompiled code ;).

Thanks

  Andrea