Get mma linux more generic #1284
Conversation
* Does not depend on helper tools like sar or top * Uses the second value which is the idle time of all cores in seconds
* Does not depend on helper function ifstat
* Does not depend on helper tool iostat
* Does not depend on helper tool ps
to be more generic
because /dev/root is a kind of virtual device
* consider all mounted disks * sum up numbers for partitions of same disk * special handling for mtdblock devices
* avoid division by zero * garbage collection for process utime data
from /etc/os-release
* runtime check if vcgencmd is available * not having this behind compile flag
![github-actions[bot]](https://avatars.githubusercontent.com/in/15368?s=60&v=4){kind=small}
contrib/mma/src/linux/mma_linux.cpp
Outdated
| proc_prev_count_ = item.count; | ||
|
|
||
| // garbage collection | ||
| for (auto it = proc_prev_map.begin(), next = it; it != proc_prev_map.end(); it = next) |
There was a problem hiding this comment.
Iterating on a map and at the same time removing certain elements can be dangerous. I am not 100% sure if your code can lead to unintended side-effects.
There was a problem hiding this comment.
The implementation follows a pattern given on Stackoverflow here: https://stackoverflow.com/a/39719450
The idea is to get the next iterator already before erasing the current entry.
The erased entries belong to old processes which are finished, means not listed in /proc/[pid] any more.
There was a problem hiding this comment.
I see and acually thought there is a std::erase_if() functionality from the library available that can be used here. But it turned out this helper method comes along with C++20 which is too "new" for our code base.
Let member variables end with `_`
Thank you this specific change! I was recording to a "data drive" and was sad when mma wasn't reporting the usage of my recording drive.. |
Description
Related issues
Cherry-pick to