Hi Maxime,

thanks for the review. I attached 2 patches.

0001-Add-emacs-sqlite3-api-package.patch

In this one I addressed your following comments:

- I changed the synopsis and the description
- Removed the make flags
- I'm using cc-for-target now
- Enabled tests

The tests are passing, and I tested it manually by evaluating the Emacs
Lisp code snippets in [1]. Everything seems to work fine.

0001-Add-emacs-sqlite3-api-package-generate-constants.patch

In this patch I did the same as above, and I generated the consts.c file
by running make in the "tools" directory. The tests are passing fine,
but when I manually test it by evaluating the Elisp snippets in [1] I
get an error that the sqlite-open-readwrite and sqlite-open-create are
not defined in Emacs. The .so file is loaded by Emacs but somehow it
does not work.

I looked at the diff between the consts.c committed in the repository
and the generated one, but I could not see any changes related to those
functions. I'm still a bit lost about this and could need some help.

But I also wonder if generating the consts.c file is actually better
than just using the checked in consts.c file. The command to generate
the file fetches a web page from the Sqlite website and builds the
consts.c file with this information. And I think this is kind of a
brittle approach. How would this approach would even work with Software
heritage?

I think my preference would be to go without generating the consts.c
file, the patch works, it uses the same file the maintainer of the
packages is using, and I believe it is less brittle. :)

What do you think?

[1] https://github.com/pekingduck/emacs-sqlite3-api#introduction