[tex-live] Problem with TL source tree (eomegadir/eover-rc1.ch)

[Thomas Esser]

> A fresh build ends in an error as well:

Not for me. Do you use the right options to preserve the timestamps as
they are on the server? Do you have acurate time on your local machine?
Do you use nfs for storing the files?

> It seems cweb.c is downloaded but with a time stamp that lets make try 
> to remake this read-only file. So I touched cweb.c and went ahead. (the 

That is a bug. cweb.c should not be there. The file usually gets removed
by a "make clean".

> same happened to me with source/TeX/texk/cjkutils/hbf2gf/hbf2gf.c)

hbf2gf.c gets removed by "make distclean". It is a bug that the file
exists in the perforce repository.

> The fact that I do not have the problem now, corroborates my hunch that 
> there *is* something wrong. After all, this way of removing all files 
> and resyncing them with p4 gives you not only base files, but it also 
> gives you intermediate results (like cweb.c) that are stored in 
> perforce. So, you *cannot* be sure that the files you get are actually 

I think that we can solve this by the following "rule": never add files
to p4 before "make distclean".

> latest modifications. In fact, that I ran into this problem suggests 
> that a basic source file was adapted and may contain a problem, but 
> that nobody notices it because during a compile, the old intermediary 
> file keeps getting used.

For teTeX, I always build outside the source directory. A following
"make distclean" must remove every file, so that only directories are
left. I do check this for every teTeX release.

Maybe, for TeX Live, more checking by the developers is needed. We have
to ensure, that
  touch foo; ./configure; make; make distclean; find . -newer foo -type f
gives empty output in the find-run. That's easy enough to check...

Thomas