gil
gil copied to clipboard
boostorg
algorithm_channel_arithmetic test seems random
Actual behavior
Porting Boost 1.69 to AIX, I see the test algorithm_channel_arithmetic failing on AIX randomly (depending on the -O? option I use). Then, looking at Fed/Intel, though the test seems to succeed, I see that intermediate values are random, probably being the root cause of my issue on AIX.
With same executable (built with addition of one line of printf in file below) on Fed/Intel: ../bin.v2/libs/gil/test/channel/algorithm_channel_arithmetic.test/gcc-9.2.1/debug/visibility-hidden/algorithm_channel_arithmetic , and launching it several times, I have the following different results:
1)
TONY: BEG f.min_v_: 0l f.max_v_: 255l
TONY: BEG f.min_v_: 4294967168l f.max_v_: 127l
TONY: BEG f.min_v_: 0l f.max_v_: 65535l
TONY: BEG f.min_v_: 4294934528l f.max_v_: 32767l
TONY: BEG f.min_v_: 0l f.max_v_: 4294967295l
TONY: BEG f.min_v_: 2147483648l f.max_v_: 2147483647l
TONY: BEG f.min_v_: 168336160l f.max_v_: 168336192l <===
TONY: BEG f.min_v_: 168336176l f.max_v_: 168336208l
TONY: BEG f.min_v_: 0l f.max_v_: 255l
2)
TONY: BEG f.min_v_: 0l f.max_v_: 255l
TONY: BEG f.min_v_: 4294967168l f.max_v_: 127l
TONY: BEG f.min_v_: 0l f.max_v_: 65535l
TONY: BEG f.min_v_: 4294934528l f.max_v_: 32767l
TONY: BEG f.min_v_: 0l f.max_v_: 4294967295l
TONY: BEG f.min_v_: 2147483648l f.max_v_: 2147483647l
TONY: BEG f.min_v_: 4049837264l f.max_v_: 4049837296l <===
TONY: BEG f.min_v_: 4049837280l f.max_v_: 4049837312l
TONY: BEG f.min_v_: 0l f.max_v_: 255l
3)
TONY: BEG f.min_v_: 0l f.max_v_: 255l
TONY: BEG f.min_v_: 4294967168l f.max_v_: 127l
TONY: BEG f.min_v_: 0l f.max_v_: 65535l
TONY: BEG f.min_v_: 4294934528l f.max_v_: 32767l
TONY: BEG f.min_v_: 0l f.max_v_: 4294967295l
TONY: BEG f.min_v_: 2147483648l f.max_v_: 2147483647l
TONY: BEG f.min_v_: 3091145712l f.max_v_: 3091145744l <===
TONY: BEG f.min_v_: 3091145728l f.max_v_: 3091145760l
TONY: BEG f.min_v_: 0l f.max_v_: 255l
4)
TONY: BEG f.min_v_: 0l f.max_v_: 255l
TONY: BEG f.min_v_: 4294967168l f.max_v_: 127l
TONY: BEG f.min_v_: 0l f.max_v_: 65535l
TONY: BEG f.min_v_: 4294934528l f.max_v_: 32767l
TONY: BEG f.min_v_: 0l f.max_v_: 4294967295l
TONY: BEG f.min_v_: 2147483648l f.max_v_: 2147483647l
TONY: BEG f.min_v_: 2127298960l f.max_v_: 2127298992l <===
TONY: BEG f.min_v_: 2127298976l f.max_v_: 2127299008l
TONY: BEG f.min_v_: 0l f.max_v_: 255l
However, on AIX, when it breaks with errors, I ALWAYS have:
TONY: BEGINNING f.min_v_: 0l f.max_v_: 255l
TONY: BEGINNING f.min_v_: 4294967168l f.max_v_: 127l
TONY: BEGINNING f.min_v_: 0l f.max_v_: 65535l
TONY: BEGINNING f.min_v_: 4294934528l f.max_v_: 32767l
TONY: BEGINNING f.min_v_: 0l f.max_v_: 4294967295l
TONY: BEGINNING f.min_v_: 2147483648l f.max_v_: 2147483647l
TONY: BEGINNING f.min_v_: 4294954660l f.max_v_: 4294954656l <===
TONY: BEGINNING f.min_v_: 4294954536l f.max_v_: 4294954528l
TONY: BEGINNING f.min_v_: 0l f.max_v_: 255l
But, still on AIX, this generates errors:
+ ../bin.v2/libs/gil/test/channel/algorithm_channel_arithmetic.test/gcc-8.4.0/debug/address-model-64/visibility-hidden/algorithm_channel_arithmetic
Running 30 test cases...
../libs/gil/test/channel/algorithm_channel_arithmetic.cpp(54): error: in "packed_channel_reference<unsigned long>": check f.min_v_ > f.max_v_ has failed [ <= ]
../libs/gil/test/channel/algorithm_channel_arithmetic.cpp(58): error: in "packed_dynamic_channel_reference<unsigned long>": check v2 == v3 has failed [ != ]
../libs/gil/test/channel/algorithm_channel_arithmetic.cpp(58): error: in "packed_dynamic_channel_reference<unsigned long>": check v2 == v3 has failed [ != ]
Expected behavior
I expect the intermediate values of f.min_v_ and f.max_v_ to be the same when I run the test. I'd like someone checking that this issue does appear in the official Boost environment.
C++ Minimal Working Example
Let's add a printf line at the begining of the test_channel_arithmetic_mutable() routine of the file libs/gil/test/channel/algorithm_channel_arithmetic.cpp .
libs/gil/test/channel/algorithm_channel_arithmetic.cpp :
template <typename ChannelFixtureBase>
void test_channel_arithmetic_mutable(boost::mpl::true_)
{
using fixture_t = fixture::channel<ChannelFixtureBase>;
using channel_value_t = typename fixture_t::channel_value_t;
fixture_t f;
channel_value_t const v = f.min_v_;
printf("TONY: BEG f.min_v_: %ul f.max_v_: %ul \n", f.min_v_, f.max_v_ );
....
Environment
- Compiler version: GCC 9.3.1 20200408 (Red Hat 9.3.1-2)
- Build settings:
- Version (Git ref or
<boost/version.hpp>): v1.69
Thanks for the report. Sorry, but I only skimmed it.
printf("TONY: BEG f.min_v_: %ul f.max_v_: %ul \n", f.min_v_, f.max_v_ );
The channels can be signed too, but I believe you are aware of that.
I do no master Boost. I used %ul because my error message on AIX is talking about "unsigned long". Anyway, using %16ld, I see:
1)
TONY: BEG f.min_v_: 0 f.max_v_: 255
TONY: BEG f.min_v_: 4294967168 f.max_v_: 127
TONY: BEG f.min_v_: 0 f.max_v_: 65535
TONY: BEG f.min_v_: 4294934528 f.max_v_: 32767
TONY: BEG f.min_v_: 0 f.max_v_: 4294967295
TONY: BEG f.min_v_: 2147483648 f.max_v_: 2147483647
TONY: BEG f.min_v_: 140729274827696 f.max_v_: 140729274827728 <===
TONY: BEG f.min_v_: 140729274827712 f.max_v_: 140729274827744
TONY: BEG f.min_v_: 0 f.max_v_: 255
2)
TONY: BEG f.min_v_: 0 f.max_v_: 255
TONY: BEG f.min_v_: 4294967168 f.max_v_: 127
TONY: BEG f.min_v_: 0 f.max_v_: 65535
TONY: BEG f.min_v_: 4294934528 f.max_v_: 32767
TONY: BEG f.min_v_: 0 f.max_v_: 4294967295
TONY: BEG f.min_v_: 2147483648 f.max_v_: 2147483647
TONY: BEG f.min_v_: 140730840531376 f.max_v_: 140730840531408 <===
TONY: BEG f.min_v_: 140730840531392 f.max_v_: 140730840531424
TONY: BEG f.min_v_: 0 f.max_v_: 255
3)
TONY: BEG f.min_v_: 0 f.max_v_: 255
TONY: BEG f.min_v_: 4294967168 f.max_v_: 127
TONY: BEG f.min_v_: 0 f.max_v_: 65535
TONY: BEG f.min_v_: 4294934528 f.max_v_: 32767
TONY: BEG f.min_v_: 0 f.max_v_: 4294967295
TONY: BEG f.min_v_: 2147483648 f.max_v_: 2147483647
TONY: BEG f.min_v_: 140725078286400 f.max_v_: 140725078286432 <===
TONY: BEG f.min_v_: 140725078286416 f.max_v_: 140725078286448
TONY: BEG f.min_v_: 0 f.max_v_: 255
Moreover, I also see that the swap() routine does not swap. However, I do not master what this code is aimed to do ; just trying to guess.
TONY: BEG f.min_v_: 2147483648 f.max_v_: 2147483647
TONY: BEF swap f.min_v_: 2147483648 f.max_v_: 2147483647
TONY: AFT swap f.min_v_: 2147483647 f.max_v_: 2147483648 Swap OK.
TONY: v2 2147483647 v3 2147483647
TONY: BEG f.min_v_: 140725539387952 f.max_v_: 140725539387984
TONY: BEF swap f.min_v_: 140725539387952 f.max_v_: 140725539387984
TONY: AFT swap f.min_v_: 140725539387952 f.max_v_: 140725539387984 No swap done.
TONY: v2 140725539387952 v3 140725539387984
TONY: BEG f.min_v_: 140725539387968 f.max_v_: 140725539388000
TONY: BEF swap f.min_v_: 140725539387968 f.max_v_: 140725539388000
TONY: AFT swap f.min_v_: 140725539387968 f.max_v_: 140725539388000
Is that correct behavior ?
printf("TONY: BEF swap f.min_v_: %16ld f.max_v_: %16ld \n", f.min_v_, f.max_v_ );
BOOST_TEST(f.min_v_ <= f.max_v_);
std::swap(f.min_v_, f.max_v_);
printf("TONY: AFT swap f.min_v_: %16ld f.max_v_: %16ld \n", f.min_v_, f.max_v_ );
BOOST_TEST(f.min_v_ > f.max_v_);
TONY: BEG f.min_v_: 4294967168 f.max_v_: 127
For the f.min_v_ above, which is 8-bit value of -128, something is casting it to unsigned 32-bit integer: 2^32 - 128, what becomes 4294967168.
I'm not sure what is happening, and I can't look into details now.
OK. I'm not in a hurry. Anyway, I think that there is an issue here. Maybe it has been fixed within last versions, I don't know, but I'd like to know. Please update this issue so that I know what I have to do for AIX.
@trex58 Yes, it looks like a bug somewhere in packed and/or bit-aligned images, possibly big-endian related. I am fluent yet about internals of those types, so it is hard to tell from top of my head.
It would be interesting to see the results for the 1.73 (soon to be released).
This is not going to be forgotten issue, but it may take time until I look into it (I will likely need QEMU running big-endian Unix).
I'm now using v1.73.0. And I think I still see the same behavior.
A) Random intermediate values.
1rst launch of test
# ../bin.v2/libs/gil/test/core/channel/algorithm_channel_arithmetic.test/gcc-9/debug/threading-multi/visibility-hidden/algorithm_channel_arithmetic 2>&1 | grep BEG
TONY: BEG f.min_v_: 0 f.max_v_: 255
TONY: BEG f.min_v_: 4294967168 f.max_v_: 127
TONY: BEG f.min_v_: 0 f.max_v_: 65535
TONY: BEG f.min_v_: 4294934528 f.max_v_: 32767
TONY: BEG f.min_v_: 0 f.max_v_: 4294967295
TONY: BEG f.min_v_: 2147483648 f.max_v_: 2147483647
TONY: BEG f.min_v_: 140725939138380 f.max_v_: 140725939138376
TONY: BEG f.min_v_: 140725939138264 f.max_v_: 140725939138256
TONY: BEG f.min_v_: 0 f.max_v_: 255
TONY: BEG f.min_v_: 4294967168 f.max_v_: 127
TONY: BEG f.min_v_: 0 f.max_v_: 65535
TONY: BEG f.min_v_: 4294934528 f.max_v_: 32767
TONY: BEG f.min_v_: 0 f.max_v_: 4294967295
TONY: BEG f.min_v_: 2147483648 f.max_v_: 2147483647
TONY: BEG f.min_v_: 140725939138348 f.max_v_: 140725939138344
TONY: BEG f.min_v_: 140725939138232 f.max_v_: 140725939138224
TONY: BEG f.min_v_: 140725939138336 f.max_v_: 140725939138328
....
2nd launch of test.
TONY: BEG f.min_v_: 0 f.max_v_: 255
TONY: BEG f.min_v_: 4294967168 f.max_v_: 127
TONY: BEG f.min_v_: 0 f.max_v_: 65535
TONY: BEG f.min_v_: 4294934528 f.max_v_: 32767
TONY: BEG f.min_v_: 0 f.max_v_: 4294967295
TONY: BEG f.min_v_: 2147483648 f.max_v_: 2147483647
TONY: BEG f.min_v_: 140729678367580 f.max_v_: 140729678367576 <<<<<
TONY: BEG f.min_v_: 140729678367464 f.max_v_: 140729678367456
TONY: BEG f.min_v_: 0 f.max_v_: 255
TONY: BEG f.min_v_: 4294967168 f.max_v_: 127
TONY: BEG f.min_v_: 0 f.max_v_: 65535
TONY: BEG f.min_v_: 4294934528 f.max_v_: 32767
TONY: BEG f.min_v_: 0 f.max_v_: 4294967295
TONY: BEG f.min_v_: 2147483648 f.max_v_: 2147483647
TONY: BEG f.min_v_: 140729678367548 f.max_v_: 140729678367544
TONY: BEG f.min_v_: 140729678367432 f.max_v_: 140729678367424
TONY: BEG f.min_v_: 140729678367536 f.max_v_: 140729678367528
....
3rd launch
TONY: BEG f.min_v_: 0 f.max_v_: 255
TONY: BEG f.min_v_: 4294967168 f.max_v_: 127
TONY: BEG f.min_v_: 0 f.max_v_: 65535
TONY: BEG f.min_v_: 4294934528 f.max_v_: 32767
TONY: BEG f.min_v_: 0 f.max_v_: 4294967295
TONY: BEG f.min_v_: 2147483648 f.max_v_: 2147483647
TONY: BEG f.min_v_: 140724722884844 f.max_v_: 140724722884840
TONY: BEG f.min_v_: 140724722884728 f.max_v_: 140724722884720
TONY: BEG f.min_v_: 0 f.max_v_: 255
TONY: BEG f.min_v_: 4294967168 f.max_v_: 127
TONY: BEG f.min_v_: 0 f.max_v_: 65535
TONY: BEG f.min_v_: 4294934528 f.max_v_: 32767
TONY: BEG f.min_v_: 0 f.max_v_: 4294967295
TONY: BEG f.min_v_: 2147483648 f.max_v_: 2147483647
TONY: BEG f.min_v_: 140724722884812 f.max_v_: 140724722884808 <<<<
TONY: BEG f.min_v_: 140724722884696 f.max_v_: 140724722884688
TONY: BEG f.min_v_: 140724722884800 f.max_v_: 140724722884792
B) swap no done
# ../bin.v2/libs/gil/test/core/channel/algorithm_channel_arithmetic.test/gcc-9/debug/threading-multi/visibility-hidden/algorithm_channel_arithmetic 2>&1 | grep -v BEG
TONY: BEF swap f.min_v_: 0 f.max_v_: 255
TONY: AFT swap f.min_v_: 255 f.max_v_: 0
TONY: v2 255 v3 255
TONY: BEF swap f.min_v_: 4294967168 f.max_v_: 127
TONY: AFT swap f.min_v_: 127 f.max_v_: 4294967168
TONY: v2 127 v3 127
TONY: BEF swap f.min_v_: 0 f.max_v_: 65535
TONY: AFT swap f.min_v_: 65535 f.max_v_: 0
TONY: v2 65535 v3 65535
TONY: BEF swap f.min_v_: 4294934528 f.max_v_: 32767
TONY: AFT swap f.min_v_: 32767 f.max_v_: 4294934528
TONY: v2 32767 v3 32767
TONY: BEF swap f.min_v_: 0 f.max_v_: 4294967295
TONY: AFT swap f.min_v_: 4294967295 f.max_v_: 0
TONY: v2 4294967295 v3 4294967295
TONY: BEF swap f.min_v_: 2147483648 f.max_v_: 2147483647
TONY: AFT swap f.min_v_: 2147483647 f.max_v_: 2147483648 Swap done
TONY: v2 2147483647 v3 2147483647
TONY: BEF swap f.min_v_: 140725453353352 f.max_v_: 140725453353348
TONY: AFT swap f.min_v_: 140725453353360 f.max_v_: 140725453353356 No swap done
TONY: v2 140725453353372 v3 140725453353368
TONY: BEF swap f.min_v_: 140725453353280 f.max_v_: 140725453353272
TONY: AFT swap f.min_v_: 140725453353296 f.max_v_: 140725453353288
Please note that I see 59 warning messages saying things like:
../boost/gil/channel_algorithm.hpp:82:67: warning: conversion from ‘long unsigned int’ to ‘int’ may change value [-Wconversion]
../libs/gil/test/core/channel/algorithm_channel_arithmetic.cpp:40:14: warning: conversion from ‘int’ to ‘boost::gil::test::fixture::channel_value<unsigned char>::channel_t’ {aka ‘unsigned char’} may change value [-Wconversion]
...
../boost/gil/channel.hpp:513:115: warning: conversion from ‘int’ to ‘boost::gil::detail::packed_channel_reference_base<boost::gil::packed_channel_reference<short unsigned int, 11, 5, true>, short unsigned int, 5, true>::bitfield_t’ {aka ‘short unsigned int’} may change value [-Wconversion]