Build of 9.0 failed at R

[Malcolm MacCallum]

I am using a Dell Inspiron 13, 5000 series, running

Ubuntu Linux 18.01.

The error in building R as seen in the attached log

led to the overall build failing. I would have liked

to have an updated sage having just changed Linux

distro to Ubuntu. Advice on how to fix this welcomed!

[Frédéric Chapoton]

If you repeat "make build -j4" until this is the only failure, this does not prevent sage to work.

So if you do not need R, you can ignore that.

I have the same problem on several machine with Ubuntu 19.10

F

[Eric Gourgoulhon]

Hi Malcolm,

I have the same OS as you (Ubuntu 18.04) and I could build 9.0 (as well as 9.1.beta0) without any trouble. If I compare the file r-3.6.2.log resulting from the build of R on my system with yours, they are pretty similar until these lines in your r-3.6.2.log:

making pretty.d from pretty.c

making uncmin.d from uncmin.c
make[7]: Entering directory '/usr/local/CA/sage-9.0/local/var/tmp/sage/build/r-3.6.2/src/src/appl'
gcc -I. -I../../src/include -I../../src/include  -DHAVE_CONFIG_H   -fopenmp -fpic  -g -O2   -c integrate.c -o integrate.o
In file included from ../../src/include/R_ext/Applic.h:41:0,
                 from integrate.c:33:
../../src/include/R_ext/BLAS.h:103:55: error: expected declaration specifiers or '...' before ')' token
   const double *beta, double *y, const int *incy FCLEN);

The similar lines in my r-3.6.2.log are

making pretty.d from pretty.c
making uncmin.d from uncmin.c
make[7]: Entering directory '/home/eric/sage/9.0/local/var/tmp/sage/build/r-3.6.2/src/src/appl'
gcc -I. -I../../src/include -I../../src/include  -DHAVE_CONFIG_H   -fopenmp -fpic  -g -O2   -c integrate.c -o integrate.o
gcc -I. -I../../src/include -I../../src/include  -DHAVE_CONFIG_H   -fopenmp -fpic  -g -O2   -c interv.c -o interv.o
...

Successfully installed r-3.6.2
Deleting temporary build directory
/home/eric/sage/9.0/local/var/tmp/sage/build/r-3.6.2
Finished installing r-3.6.2.spkg

A question: are you using the system gfortran (i.e. gfortran coming from the Ubuntu package gfortran) and not Sage's built gfortran? I am using the system one.

Best regards,

Eric.

Le samedi 11 janvier 2020 16:18:37 UTC+1, Malcolm MacCallum a écrit :

[Eric Gourgoulhon]

PS: I am attaching my log file for R.

[Eric Gourgoulhon]

From the log files, a difference between your system and mine is that your Linux kernel is 5.0.0-37-generic, while mine is 4.15.0-72-generic (which is the standard for Ubuntu 18.04.3, I think). I don't know is this could be the root of the difference in the building of R.

Best regards,

Eric.

[Dima Pasechnik]

we have seen this already, it is a known bug/feature of gfortran (certain versions) found by R people.

--
You received this message because you are subscribed to the Google Groups "sage-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email to sage-devel+...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/sage-devel/3175f829-2d80-4bbb-b1ef-800052dda26c%40googlegroups.com.

[Eric Gourgoulhon]

Le dimanche 12 janvier 2020 16:40:21 UTC+1, Dima Pasechnik a écrit :

we have seen this already, it is a known bug/feature of gfortran (certain versions) found by R people.

Ah yes:

https://developer.r-project.org/Blog/public/2019/05/15/gfortran-issues-with-lapack/

In my case, gfortran --version returns

GNU Fortran (Ubuntu 7.4.0-1ubuntu1~18.04.1) 7.4.0

Eric.

On Sun, 12 Jan 2020, 15:36 Eric Gourgoulhon, <egourg...@gmail.com> wrote:

From the log files, a difference between your system and mine is that your Linux kernel is 5.0.0-37-generic, while mine is 4.15.0-72-generic (which is the standard for Ubuntu 18.04.3, I think). I don't know is this could be the root of the difference in the building of R.

Best regards,

Eric.

Le dimanche 12 janvier 2020 15:19:59 UTC+1, Eric Gourgoulhon a écrit :

PS: I am attaching my log file for R.

--
You received this message because you are subscribed to the Google Groups "sage-devel" group.

To unsubscribe from this group and stop receiving emails from it, send an email to sage-...@googlegroups.com.

[Malcolm MacCallum]

Thanks to all those who replied, especially Eric.

It seems clear the issue is/was with gfortran. I had not installed the system gfortran so I was

using the sage one. I now have the system one (7.4.0, same as Eric). I'm not entirely sure how

to ensure sage uses the system version (in principle I view the self-containedness of

sage as one of its major virtues). Using

make build -j4

has built a sage 9.0 that starts up OK (not fully tested yet), and since I don't use R

that should be fine.