Hi Bruno,
> Also, please don't drop comments that are present in the original code, such
> as:
>
> # Replace NMD, so as to remove redundant "$(MKDIR_P) '.'" invocations.
> # The logic is similar to how we define gl_source_base_prefix.
Sure. I remember seeing that comment in the shell script when emitting
lib/Makefile.am and tests/Makefile.am. I'll try to add it in similar
positions so it is easy to cross reference.
> Recall that ultimately we want to drop the gnulib-tool shell implementation.
That is exciting. As a small progress update, these changes (aside
from the issues you have mentioned) are enough to bootstrap Coreutils.
Here is the time of each bootstrap in a clean directory:
# gnulib-tool
$ time ./bootstrap
...
./bootstrap: done. Now you can run './configure'.
real 2m31.575s
user 0m59.316s
sys 1m45.466s
# gnulib-tool.py
$ time ./bootstrap
...
./bootstrap: done. Now you can run './configure'.
real 0m34.951s
user 0m16.930s
sys 0m5.395s
I can then ./configure && make all && make check and everything works
as normal which is a good sign. I would advise against this for now
until I can fix a few more issues of course, but just to show things
are moving in the right direction.
For example, after a day of testing I get this:
$ find /tmp -name 'tmp*' 2> /dev/null | wc -l
141
As a result of the exit() in GLImport I assume. :)
> The only patch that I've applied from you so far, today, is 0005.
> 0001, 0002, 0003, 0004 all need to be revisited.
Thanks for all the feedback. I'll get around to responding with fixed
patches sometime today or tomorrow.
Collin
