Re: New branch 201906

[Katsumi Yamaoka <yamaoka@xxxxxxx>]

>> In [emacs-w3m:13457]
>> On Mon, 17 Jun 2019 16:12:26 +0900, Katsumi Yamaoka wrote:
>>> Tested and tweaked the code so that `make slow' with Emacs 25.1,
>>> 26.1, 26.2.90, and 27.0.50 may issue no warning.

>> Sorry.  Not `make slow' but `make very-slow' that compiles the
>> .el files one by one individually.

`one by one individually' means to run Emacs with -q and some
arguments (that includes --eval '(push "." load-path)'), byte-
compile *an* .el file, and exit Emacs.  This way is applied for
all the .el files.  Moreover, `very-slow' does it in two passes,
one is that .el files other than the one beeing byte-compiled
are not byte-compiled (there is no .elc file), and the other is
that .el files other than the one beeing byte-compiled have been
all byte-compiled (only the .elc file for the one being compiled
is absent).  If w3m-foo.el uses the features that cl-lib provides
whereas it does not have (eval-when-compile (require 'cl-lib)),
it will be compiled successfully in the first pass if w3m-foo.el
requires w3m-bar and it has (eval-when-compile (require 'cl-lib)),
however it will fail in the second pass since eval-when-compile
lines are not in the w3m-bar.elc file.  Then we get to know that
w3m-foo.el needs to have (eval-when-compile (require 'cl-lib)).
The second pass of `very-slow' is significant, the first pass is
existent for mainly byte-compiling all the .el files.  What we
aime at is to ensure any .el file to be byte-comipled with no
constraint.  `very-slow' takes a long time, so it is named so.
`slow' is similar to the first pass of `very-slow', just `make'
does byte-compiling all the .el files in one Emacs process.

Regards,