Re: make check when cross compiling

[Harry Thijssen]

Ok

Then the problem might be that the file in tests/data is sack.exe and not sack.  Also I have seen errors for pspp which I guess should be pspp.exe.

Might there be a problem?

Have fun

2016-03-09 18:11 GMT+01:00 Ben Pfaff <address@hidden>:

On Wed, Mar 09, 2016 at 01:57:49PM +0100, Harry Thijssen wrote:
> Hi
>
> I had a look to find out why all tests fail when cross compiling.
>
> If I look at atconfig.in I see:
>
> # -*- shell-script -*-
>
> # Variables used internally by the testsuite.
> EXEEXT='.exe'
> GNM_READ_SUPPORT='yes'
> ODF_READ_SUPPORT='yes'
> .....
>
>
> if I look in the logfiles from the tests I see:
>
> #                             -*- compilation -*-
> 93. sys-file-reader.at:1908: testing missing string continuation record ...
> ./sys-file-reader.at:1926: sack --$variant sys-file.sack > sys-file.sav
> --- /dev/null    2016-03-08 20:27:55.663840673 +0100
> +++
> /home/harry/pspp-master-20160308/pspp-0.9.0-gc50266/tests/testsuite.dir/at-groups/93/stderr
> 2016-03-08 22:01:34.369722134 +0100
> @@ -0,0 +1 @@
> +/home/harry/pspp-master-20160308/pspp-0.9.0-gc50266/tests/testsuite.dir/at-groups/93/test-source:
> line 25: sack: command not found
> ./sys-file-reader.at:1926: exit code was 127, expected 0
> 93. sys-file-reader.at:1908: 93. missing string continuation record (
> sys-file-reader.at:1908): FAILED (sys-file-reader.at:1926)
>
> Shouldn't it be sack.exe instead of sack?

No, sack runs on the machine where the build occurs, not on the machine
that the build targets.  In other words, sack is a build tool.