Make RTLD_LOCAL work correctly for recursively loaded DSOs#18924
Conversation
sbc100
force-pushed
the
rtld_local2
branch
2 times, most recently
from
3dad6f6 to
ee8da4b
Compare
src/library_dylink.js
Outdated
|
|
||
| if (typeof result == 'function') { | ||
| // Workaround for closure compiler error if f.stub is accessed directly | ||
| // on the line below. |
There was a problem hiding this comment.
Weird error... what's the error message, if you still have it?
There was a problem hiding this comment.
Closure says that Function does not have a stub attribute. Which is true for most functions, but we sometimes at them, which I guess is not something closure likes in general (attributes that might or might not exist). Should I add more detail to the comment?
There was a problem hiding this comment.
No need, thanks, I was just curious.
There was a problem hiding this comment.
I think the problem is that closure treat the if (typeof result == 'function') type refinement and will then assume that result to be of type Function.
| shutil.move(outfile, 'liblib.so') | ||
| cmd = [compiler_for(filename), filename, '-o', outfile] + self.get_emcc_args() | ||
| if emcc_args: | ||
| cmd += emcc_args |
There was a problem hiding this comment.
Why do we need this in addition to get_emcc_args() on the prevoius line?
There was a problem hiding this comment.
I used it below on the line self.build_dlfcn_lib('liba.c', outfile='liba.so', emcc_args=['libb.so'])
Here i want to set libb.so as an argument to compile step, but I don't want to add it to all the following compile steps in the test.
We have several functions that both take emcc_args and honor the global emcc_args. Sometimes it makes more sense to use one over the other I think.
There was a problem hiding this comment.
I see, thanks, I was just confused by the name then.
| shutil.move(outfile, 'liblib.so') | ||
| cmd = [compiler_for(filename), filename, '-o', outfile] + self.get_emcc_args() | ||
| if emcc_args: | ||
| cmd += emcc_args |
There was a problem hiding this comment.
I see, thanks, I was just confused by the name then.
When dlopen is first called with `RTLD_LOCAL` a new local scope is created. All symbols from the DSO being loaded and any of its transitive dependencies get added to this scrope (instead of the global scope).
…n-core#18924) When dlopen is first called with `RTLD_LOCAL` a new local scope is created. All symbols from the DSO being loaded and any of its transitive dependencies get added to this scrope (instead of the global scope).
…n-core#18924) When dlopen is first called with `RTLD_LOCAL` a new local scope is created. All symbols from the DSO being loaded and any of its transitive dependencies get added to this scrope (instead of the global scope).
2 participants
When dlopen is first called with
RTLD_LOCALa new local scope is created. All symbols from the DSO being loaded and any of its transitive dependencies get added to this scrope (instead of the global scope).