Re: make distclean and quail

[Juanma Barranquero <lekktu@gmail.com>]

On Sun, Aug 11, 2013 at 7:17 PM, Eli Zaretskii <eliz@gnu.org> wrote:

> So you are saying that _all_ of the Makefile's set srcdir to its
> correct value

Yes, all Makefiles have srcdir = /c/Devel/emacs/repo/trunk [etc],

> and yet the commands issued by Make have "Devel"
> downcased, is that right?

The "Entering directory" / "Leaving directory" messages are correctly
cased, as are many commands:

  /c/Devel/emacs/build/bin/makeinfo --force
-I/c/Devel/emacs/repo/trunk/doc/misc/../emacs [etc]

  gcc  -std=gnu99 -mtune=pentium4 [deleted warnings stuff] -I.
-I/c/Devel/emacs/repo/trunk/nt  -Lc:/Devel/emacs/lib -mtune=pentium4
-Ic:/Devel/emacs/include -DUSE_CRT_DLL=1 -I
/c/Devel/emacs/repo/trunk/nt/inc -O0 -g3
/c/Devel/emacs/repo/trunk/nt/runemacs.c

However "shell" commands are not:

  [ -r "/c/devel/emacs/repo/trunk/src/config.in" ] || ( cd
/c/devel/emacs/repo/trunk && autoheader )

  if test -f "/c/devel/emacs/repo/trunk/src/.gdbinit"; then rm -f
./.gdbinit.save; \

  make[1]: Entering directory `/c/Devel/emacs/repo/trunk/leim'
  SRC = /c/devel/emacs/repo/trunk/leim
  PWD = /c/Devel/emacs/repo/trunk/leim
  [ `cd /c/devel/emacs/repo/trunk/leim && /bin/pwd` != `/bin/pwd` ] &&
rm -rf quail

    CONFIG_SITE=/c/Devel/emacs/repo/trunk/nt/mingw-cfg.site
./config.status --recheck;  \
    CONFIG_SITE=/c/Devel/emacs/repo/trunk/nt/mingw-cfg.site
/c/devel/emacs/repo/trunk/configure --cache-file=/dev/null; \

> That kinda blames Make and it alone,
> because it is Make who expands variables when it invokes commands.
> Did you try to install a newer snapshot of MSYS Make?

I'm using the one I installed originally from here:
https://sourceforge.net/projects/mingwbuilds/files/external-binary-packages/

  C:\...\repo> \bin\MinGW\msys\1.0\bin\make.exe --version
  GNU Make 3.82.90
  Built for i686-pc-msys

> And you also say that this happens only sometimes?

Yes. Perhaps when a reconfiguration is triggered, though I'm not sure
because it does not happen so often as to clearly establish a pattern.

> Is it possible
> that you invoke the bootstrap a bit differently when the problem
> happens, like with different environment variables or from a different
> shell?

It is highly unlikely, because I always run bootstrap with a TCC
alias, which invokes "sh.exe --login -i", which reads the same init
file, which sets the path to a limited set of directories, always the
same, and I haven't really reinstalled or changed anything in there
recently, and certainly not between my last and next-to-last reboots
(from 2013/08/08 and 2013/08/11).

> Or maybe you have more than one Make which could come into
> play somehow?

There's anothe make in the path, the one from GnuWin32

  C:\...\repo> \bin\Gnu\bin\make.exe --version
  GNU Make 3.81

but I don't see why would it be run, when the other one comes before
in the path, and in fact running the very same shell command that I
use for bootstrap, is the MSYS one which responds:

  C:\...\repo> mdo make --version
  GNU Make 3.82.90
  Built for i686-pc-msys

and mboot, the command that I use to bootstrap, is defined to run "mdo
make bootstrap".

> FWIW, this never happened to me, but then I don't think I've built
> packages using MSYS in a directory that has mixed-case letters, they
> are always lower-case here.

Well, I certainly can rename C:/Devel to C:/devel and not to worry
again, but it is still quite puzzling why does it happens only part of
the time.

  J