Bug#918329: apriltag ftbfs on release architectures

[Adrian Bunk]

Control: reassign -1 gcc-8
Control: affects -1 src:apriltag

On Sat, Jan 05, 2019 at 08:44:32AM +0100, Matthias Klose wrote:
> Package: src:apriltag
> Version: 0.10.0-1
> Severity: serious
> Tags: sid buster
> 
> apriltag ftbfs on release architectures, where it built before.  Looks like the
> build system is exaggerating optimization flags, using -O4 nonsense.

-O4 is treated as -O3, so that is harmless.

The actual bug is in gcc, and also present with -O2.

Compiling a not too complicated C file that assigns different buildtime 
integer constants to all 65k members of an array is not very fast on amd64
(this is Coffee Lake at 4.7 GHz):

$ time gcc -c -O2 tagCircle49h12.c

real    0m51.891s
user    0m51.775s
sys     0m0.116s
$ 


But gcc speed on some architectures like ppc64el or arm64 is really pathetic:

(sid_ppc64el-dchroot)bunk at plummer:~/build/apriltag-0.10.0$ time gcc -c -O2 tagCircle49h12.c

real    150m38.971s
user    149m18.332s
sys     0m57.941s
(sid_ppc64el-dchroot)bunk at plummer:~/build/apriltag-0.10.0$ 

(sid_arm64-dchroot)bunk at amdahl:~/build/apriltag-0.10.0$ time gcc -c -O2 tagCircle49h12.c

real    216m20.266s
user    215m47.712s
sys     0m2.172s
(sid_arm64-dchroot)bunk at amdahl:~/build/apriltag-0.10.0$ 


cu
Adrian

-- 

       "Is there not promise of rain?" Ling Tan asked suddenly out
        of the darkness. There had been need of rain for many days.
       "Only a promise," Lao Er said.
                                       Pearl S. Buck - Dragon Seed