file installation fix

[gerard@…]

apparently running "touch *" in the file source dir. may fix the problem that's caused by automake, so the sed would be unneeded.

I haven't tested this, so somebody please do

[gerard@…]

<mca> Gerard: file-3.3*.. touch aclocal.m4 configure Makefile.in in that order to stop the regeneration

[gerard@…]

It fixes regeneration but it will call 'autoheader' in the file-3.37 directory after the compilation. Running 'autoheader' will fail. I'm not sure what is the best thing to do now. Let's discuss on lfs-dev and see what happens

[gerard@…]

never mind, figured it out. add stamp-in.h to the list of files to 'touch'

[gerard@…]

I've tried to run a diff on the generated files when they are compiled with the sed method and with the touch method. The /usr/bin/file binary differs, perhaps there's a timestamp or whatever in it.

So, we're going to have to find a better way to test and make sure the program works in the exact same way. If not, find out which one is better.

[gimli@…]

root:/usr/src# diff file-3.36 file-3.36a |egrep ^{diff diff file-3.36/Makefile file-3.36a/Makefile diff file-3.36/Makefile.in file-3.36a/Makefile.in diff file-3.36/aclocal.m4 file-3.36a/aclocal.m4 diff file-3.36/config.h file-3.36a/config.h diff file-3.36/config.log file-3.36a/config.log diff file-3.36/config.status file-3.36a/config.status diff file-3.36/configure file-3.36a/configure diff file-3.36/readelf.h file-3.36a/readelf.h}

These are the files that differ. Only readelf.h and config.h affect the final binary. readelf.h was changed by the sed of course, that leaves config.h. The only changes I see are typedef changes. So, I think there's nothing to worry about so just running touch * produces a more or less identical binary, at least it does the same.

[gerard@…]

Good enough for me. Let's do it.

[markh@…]

forgot to put the comment in..

checked during a CVS test build. This works with file-3.37 which is now in cvs. Closing