On 29 January 2015 at 15:30, Lubomir I. Ivanov <neolit...@gmail.com> wrote: > On 29 January 2015 at 04:59, Thiago Macieira <thi...@macieira.org> wrote: >> On Wednesday 28 January 2015 11:04:21 Linus Torvalds wrote: >>> The proper way to get the git version is to just do >>> >>> # get raw SHA1 >>> git rev-parse HEAD >>> >>> # get "description" of it >>> git describe --tags --abbrev=12 >>> >>> but actually accessing the .git/HEAD file directly is very wrong. But >>> I have no idea how to do git commands in the *.pri file, and handling >>> failure gracefully (in case it's not a git repository etc). >> >> You're right, but this implies always running the rule. That implies Make >> must >> find that a given file is not up to date, otherwise it won't run the rule. In >> turn, it means the not-up-to--date status cascades down to the binary and >> Make >> will recompile and relink, even if nothing changed. >> >> I don't know of a way to ask Make to always run some commands and inspect >> file >> contents before deciding what is up to date and what isn't. The closest I can >> think of is to touch or not touch another file, but in a parallel build Make >> may have already inspected that file and decided it was up-to-date. >> >> Do you know of any tricks I'm missing? > > this is what i would do: > ----------------- > > all: main.exe > > .PHONY: persist > > version.h: persist > cat $@ 2> /dev/null || git rev-parse HEAD > $@ > git rev-parse HEAD > $@.tmp > git diff --quiet $@ $@.tmp || cp $@.tmp $@
"diff -q" should be used instead of "git diff": because version.h should not be tracked by git and "git diff" is not reliable if the file is not tracked (i think). at least it returns no output with msys-git no matter the files (.h / .h.tmp) being differ or not. lubomir -- _______________________________________________ subsurface mailing list subsurface@subsurface-divelog.org http://lists.subsurface-divelog.org/cgi-bin/mailman/listinfo/subsurface
