Closed devurandom closed 7 years ago
Funny, it doesn't fail here. What is your shell? I am guessing we are dealing with a bashism.
I am still being shocked at how inconsistent makefiles are for this kind of stuff. It is a big QA in any case. Fix coming when I get to work.
OK, that should fix the issue. I also did some clean up and fixed parallel make (removing emake -j1
) while I was at it.
I have seen this on my gentoo system. (Working with Sage from source).
Should we fix this in Sage (package gap_packages) too?
You are welcome to adopt it if you want. The commit does a lot more than fixing this particular issue, you may want to be a bit selective. I can review the ticket if you like.
I must say I don't understand the issue, why cd leon make
apparently works on other systems, but not on gentoo. (and there is a missing makefile in leon/
?!)
That's a concern. cd leon make
is certainly not portable. There are at least 2 QA problems with that statement - I don't understand that it works at all on a lot of systems.
There should be a Makefile.in
in src/leon
and configure
in src/leon
will generate your makefile as usual. Left on its own the top makefile will run configure
in src/leon
but there may be a few dependency issue to carefully look at in that top makefile.
In Gentoo the approach is to run src/leon/configure
in the configure phase and not let the Makefile do it during building.
I should add that I didn't experience the problem on my own gentoo systems. I suspect the problem is linked to what is set as the default sh
or possibly your default shell.
my bin/sh
is a link to bash
. IMHO every shell I have is actually bash
.
Well then I have currently no ideas what is causing the trouble on some system. I just know that I am not surprised at all that it fails. Ticket on trac?
here you are: https://trac.sagemath.org/ticket/22116
build.log.zip