Add environment variable for tracking child processes #84
Conversation
|
The #80 issue isn't really related; these are two entirely different use cases:
Correct. This is an user error. We could probably try to detect that and print an error or something, but it's not critical. |
| pid_t pid = fork(); | ||
| if( pid == 0 ) { | ||
| // Child | ||
| if (execl("/usr/bin/ls", "/usr/bin/ls", NULL) == -1) { |
There was a problem hiding this comment.
Could you use the basic.c program here instead and in the integration test verify that:
- the file for the parent is created and contains the allocations we expect,
- the file for the child is created and contains the allocations we expect (you can just grab the checks from
#[test]for thebasic.c, move it into a separate functions, and then call it in two places)
There was a problem hiding this comment.
Good idea, I changed the test to incorporate the basic program.
docs/src/configuration.md
Outdated
| If set to `1`, bytehound will also track the memory allocations of child processes spawned by the | ||
| profiled process. |
There was a problem hiding this comment.
Could you explicitly also document that:
- this only applies to child processes which are
forkd andexecd, - that the
MEMORY_PROFILER_OUTPUTneeds to be set with the appropriate placeholders so that the filenames of both the parent and the child are different
|
LGTM; thanks! |
Hi! We want to add
bytehoundto therustc-perfsuite for profiling the Rust compiler. However, the compiler is usually invoked throughcargo, which executes manyrustcchild processes underneath. In theory this could be resolved with someRUSTC_WRAPPERhacks, but I thought that it would be better if we could directly teachbytehoundto track child processes.I'm not sure what is the interaction of the added flag with
MEMORY_PROFILER_OUTPUT, if it's set to a single, fixed filepath without placeholders, then the profile could be overwritten by the child processes? But that's probably user error.