Magic Lantern Forum

Developing Magic Lantern => General Development => Topic started by: bischofftep on April 18, 2014, 05:56:35 PM

Title: Benchmark Stub Test failures?
Post by: bischofftep on April 18, 2014, 05:56:35 PM
Hello, all:

Got a development environment up and running under VirtualBox and everything compiles fine. Several things aren't working as they do under nightly builds, though, the most noticeable being that the Benchmark->Stub Test reveals a LOT of [FAIL] markers. For example:
Code: [Select]
[Pass] UNCACHEABLE(p) => 0x42000068
       _shoot_free(p)
       m11 = MALLOC_FREE_MEMORY => 0x411b8
       m12 = GetFreeMemForAllocateMemory() => 0x2c865c
[Pass] ABS(m01-m11) => 0x0
[Pass] ABS(m02-m12) => 0x0
[FAIL] ABS(m0-m2) => 0x1518
[FAIL] ABS((m0-m1) - 256*1024) => 0x14f0
[FAIL] ABS(m0-m2) => 0x14fc
[FAIL] ABS((m0-m1) - 256*1024) => 0x14f0
[FAIL] ABS(m0-m2) => 0x14fc

followed by a whole string of [FAIL] just like the ones shown.

My build pattern is:
/magic-lantern/make
/magic-lantern/platform/7D.203/make install

That's it... any ideas what might be causing this?
Title: Re: Benchmark Stub Test failures?
Post by: nanomad on April 19, 2014, 03:09:05 PM
What compiler version are you using? (of gcc-arm-none-eabi)
Title: Re: Benchmark Stub Test failures?
Post by: bischofftep on April 22, 2014, 03:10:09 PM
gcc-arm-none-eabi-4_8-2014q1
Title: Re: Benchmark Stub Test failures?
Post by: a1ex on April 22, 2014, 03:24:55 PM
Probably the threshold is a bit too tight (it's just an approximation to check if malloc/free seems to reserve the requested memory and give it back). If some other code is using the memory allocators in parallel with the tests, they may fail.
Title: Re: Benchmark Stub Test failures?
Post by: bischofftep on April 22, 2014, 04:32:03 PM
Probably the threshold is a bit too tight (it's just an approximation to check if malloc/free seems to reserve the requested memory and give it back). If some other code is using the memory allocators in parallel with the tests, they may fail.

Interesting... I went ahead and dropped back to the 2013_q4 toolchain just to be sure it wasn't something else going on, same problems.

Nightly builds do NOT have this problem, just my own builds. Frustrating, and makes me worry that my development environment isn't working after all.