On Mon, Feb 14, 2022 at 08:06:06PM +0100, pertu...@free.fr wrote: > On Mon, Feb 14, 2022 at 03:21:59PM +0000, Werner LEMBERG wrote: > > > > > I reverted that change, and instead commited a fix using a > > > generation of tp_api.texi in autogen.sh, based on your earlier > > > proposal. Please check that it is ok... > > > > It works for me, too, thanks :-) > > > > Another thing: After `./autogen.sh && ./configure && make` in the git > > repository, doing `git diff` is longer than 15000 lines due to updated > > `.po` files and file `tp/Texinfo/XS/gnulib/lib/unistd.h` removed. Is > > there any chance that this could be improved? > > `tp/Texinfo/XS/gnulib/lib/unistd.h` is likely to be a file than should > not be in the VCS if it can be added/removed based on configure output. > There is a unistd.in.h file, so I removed the file unistd.h > automatically generated file. > > > The normal expectation > > is that there are zero differences after calling `make`. > > We synchronize po files from time to time, but not everytime they > change, as they change each time any po message line changes, which is > very often. > > I just updated them, but this will only hold for some time...
I remember having the problem with .po files muddying diffs a long time ago, but not for years. I have suspicions that the fact that po/texinfo.pot recently starting being tracked in git is responsible. I tried making a few small changes to the messages to see if the po files changed. Running "make" normally didn't lead to the po files changing, so I tried autogen.sh followed by configure and make. This didn't produce any changes either. There is probably some way of triggering the situation. If this becomes an ongoing problem we could try untracking po/texinfo.pot as a possible solution.
