[PATCH v2] api: system: allow odp_system_meminfo_t.total_overhead to be zero #2184
Conversation
odpbuild
changed the title
test/validation/api/system/system.c
Outdated
| @@ -277,7 +277,8 @@ static void system_test_meminfo(void) | |||
| CU_ASSERT(ret == ret_0); | |||
| CU_ASSERT(info_0.total_mapped == info.total_mapped); | |||
| CU_ASSERT(info_0.total_used == info.total_used); | |||
| CU_ASSERT(info_0.total_overhead == info.total_overhead); | |||
| if (info_0.total_overhead) | |||
There was a problem hiding this comment.
The condition does not seem necessary. Even if total_overhead is not supported, its value (zero in that case) should be the same in the two calls.
There was a problem hiding this comment.
What I was thinking here was that an implementation could still fill total_overhead when it has to read the per memory block info anyways. But now thinking about it, perhaps it's not the cleanest if the field value depends on other parameters.
There was a problem hiding this comment.
The "zero if not supported by the implementation" text in the API makes me think that if total overhead count is supported, then it is returned in all calls. If it might be zero also (e.g.) when the return value is less than the num parameter, then I think the API probably should mention it.
| @@ -1,6 +1,6 @@ | |||
| /* SPDX-License-Identifier: BSD-3-Clause | |||
There was a problem hiding this comment.
Typo in commit message: debuggin.
For some implementations it may not be always feasible to fill odp_system_meminfo_t.total_overhead value. As this field is only intended for debugging purposes, allow implementations to return zero in this field. Signed-off-by: Matias Elo <[email protected]> Reviewed-by: Janne Peltonen <[email protected]>
MatiasElo
force-pushed
the
api/system-meminfo
branch
from
0a8a405 to
c9433e1
Compare
odpbuild
changed the title
V2: