On Tue, 2018-07-17 at 12:35 -0700, Adam Williamson wrote: > On Fri, 2018-07-06 at 13:19 -0700, Adam Williamson wrote: > > On Fri, 2018-07-06 at 19:25 +0200, Daniel Kiper wrote: > > > On Wed, Jul 04, 2018 at 10:08:53AM -0700, Adam Williamson wrote: > > > > gentpl.py is python2/3-agnostic, but there's no way to cause it > > > > to be run with any interpreter other than 'python', it's just > > > > hard-coded into Makefile.common that way. Adjust that to allow > > > > a make variable PYTHONBIN to be set to the desired interpreter. > > > > This will make it easier in situations where we specifically > > > > want to build with 'python2' or 'python3' or whatever. > > > > > > > > Signed-off-by: Adam Williamson <awill...@redhat.com> > > > > > > Thanks for the patch. However, I think that the configure should find > > > correct python binary and set PYTHON variable (instead of PYTHONBIN) > > > in the Makefile (good example is BUILD_CC variable in Makefile.in). > > > > That's possible, but it depends what you mean by "correct", doesn't it? > > There could be many python interpreters installed on a system; which > > are we to assume is "correct"? > > > > We could make it configurable with some sort of default heuristic, I > > guess, I was just going for a simple patch approximately in line with > > what was done for autogen.sh for now. (And I wouldn't want to reinvent > > python interpreter discovery, which has been invented enough times > > already; if we were to go that route it'd probably make sense to use an > > existing autoconf extension or something). > > So I've just been looking into this, and it's actually very easy to do > using AM_PATH_PYTHON, provided by autoconf. However, there's a rather > funny wrinkle for grub's particular build process. AM_PATH_PYTHON has a > documented behaviour where, if the 'PYTHON' env var is set when > autoconf is run, it will *only* consider the value that env var is set > to as a candidate for the Python interpreter - it won't do its usual > attempt to discover one. And grub's autogen.sh already uses the PYTHON > env var - setting it if it was not set already - and then calls > autoreconf! > > So the practical effect of just applying this patch would be that only > 'python', or whatever the env var PYTHON is set to when running > autogen.sh, would be considered, when using autogen.sh. All the > 'discovery' bits of AM_PATH_PYTHON wouldn't be used: if you just call > 'autogen.sh' without setting PYTHON, but your Python binary is called > 'python3' or 'python2.7' or whatever, it won't work. > > This seems a bit funny, but I'm not sure if it's really a > *problem*...because autogen.sh *itself* will only work if whatever > value PYTHON is set to is a working Python interpreter, anyway. We > could in fact argue that the consequence is quite nice as it makes > everything sort of consistent: whatever PYTHON is set to will be the > interpreter used both by autogen.sh and for running gentpl.py. But I > figured I should point it out. > > With that in mind, I'm attaching the updated patch in git format. > Thanks!
Ping? I never received any follow up to this. -- Adam Williamson Fedora QA Community Monkey IRC: adamw | Twitter: AdamW_Fedora | XMPP: adamw AT happyassassin . net http://www.happyassassin.net _______________________________________________ Grub-devel mailing list Grub-devel@gnu.org https://lists.gnu.org/mailman/listinfo/grub-devel