make check failed on gdimagestringft_bbox
Issue #120
new
I have freetype 2.5.2, if it's relevant.
FAIL: gdimagestringft/gdimagestringft_bbox
1 of 91 tests failed Please report to ...
The system is 64-bit linux with eglibc 2.18. libgd is 2.1.0
Comments (3)
-
-
We could add a specifc test or skip it for your platform. Do you have a way to identify it in CMake and/or configure?
-
And...this test also fails on my MIPS64 platform (Octeon-II, no FPU, glibc-2.16-octeon, gcc-4.7.0-octeon, and freetype-2.5.5, built with "-O2 -mabi=64 -march=octeon -msoft-float"). Test failure message is exactly the same; gd version is 2.1.1 release.
In this case "double" is 64-bits long, so lack of precision doesn't appear to be the issue. Maybe this is just some x86-specific quirk of floating-point precision that the test is relying on?
- Log in to comment
This also fails on a custom BeyondLFS build, using ARMv7 (Cortex-A9), glibc-2.19, gcc-4.8.3, and freetype-2.5.3:
(491, 364) (613, 313) (602, 288) (481, 338) expected, but (491, 364) (613, 313) (602, 288) (480, 339)
FAIL: gdimagestringft/gdimagestringft_bbox
Note that if this is down to a rounding error, it could be because this particular ARM platform does not support double-precision floating point (type "double" is still a 32-bit float).