fix: Check for existence of_ SC_AVPHYS_PAGES to fix MacOS compilation. #3207
Conversation
Codecov ReportAll modified and coverable lines are covered by tests ✅
Additional details and impacted files@@ Coverage Diff @@
## develop #3207 +/- ##
========================================
Coverage 55.75% 55.75%
========================================
Files 1041 1041
Lines 88534 88534
========================================
Hits 49358 49358
Misses 39176 39176 ☔ View full report in Codecov by Sentry. |
|
Time to use the GitHub M1 runners? |
I found this but it's kind of old and I have not tried it. |
I don't have a MAC so it would be difficult to test alternatives. |
| m_availableMemory = sysconf( _SC_AVPHYS_PAGES ) * sysconf( _SC_PAGESIZE ); | ||
| #else | ||
| GEOS_WARNING( "Unknown device avaialable memory size getter for this system." ); |
There was a problem hiding this comment.
| GEOS_WARNING( "Unknown device avaialable memory size getter for this system." ); | |
| GEOS_WARNING( "Unknown device available memory size getter for this system." ); |
| @@ -27,8 +27,13 @@ MemoryInfos::MemoryInfos( umpire::MemoryResourceTraits::resource_type resourceTy | |||
| case umpire::MemoryResourceTraits::resource_type::pinned: | |||
| #if defined( _SC_PHYS_PAGES ) && defined( _SC_PAGESIZE ) | |||
| m_totalMemory = sysconf( _SC_PHYS_PAGES ) * sysconf( _SC_PAGESIZE ); | |||
| #if defined(_SC_AVPHYS_PAGES) | |||
| m_availableMemory = sysconf( _SC_AVPHYS_PAGES ) * sysconf( _SC_PAGESIZE ); | |||
| #else | |||
There was a problem hiding this comment.
Would something like this be a way to find the available memory on MaxOS? Unfortunately, I we don't have any MacOS to test that here.
| #else | |
| #elif /* is compiling for MacOS */ | |
| { | |
| uint64_t mem; | |
| size_t len = sizeof( mem ); | |
| sysctlbyname( "hw.memsize", &mem, &len, nullptr, 0 ); | |
| m_availableMemory= mem / sysconf(_SC_PAGE_SIZE); | |
| } |
Should we regroup the system-dependant component like this in a namespace ?
is it now available? I thought that we could not find any M1 runner but if they now have one we really should. It's true that we don't really need to support MacOS but a lot of us work on macbooks so it would be convenient to have a CI job to cover it. |
rrsettgast
changed the title
#3207) * check for existence of_ SC_AVPHYS_PAGES --------- Co-authored-by: Randolph Settgast <[email protected]>
The merge of PR has broken compilation on my mac laptop. I have patched it like this.