901282 - dlsym(dlopen) should return __wrap_dlopen instead of system dlopen

Status: RESOLVED WONTFIX

(Core :: mozglue, defect)

Description

[George Wright (:gw280) (needinfo me!)]

This behaviour currently breaks apitrace, because it calls dlopen by getting a function pointer from dlsym(RTLD_DEFAULT, "dlopen"), which returns system dlopen. It then dies later on because it calls dlsym directly, which calls into __wrap_dlsym and has an incompatible handle type to system dlopen.

Comment 1

[George Wright (:gw280) (needinfo me!)]

Attachment: 0001-Bug-901282-dlsym-dlopen-should-return-__wrap_dlopen-.patch

Comment 2

[Mike Hommey [:glandium]]

Comment on attachment 785442 [details] [diff] [review]
0001-Bug-901282-dlsym-dlopen-should-return-__wrap_dlopen-.patch

Review of attachment 785442 [details] [diff] [review]:
-----------------------------------------------------------------

You can probably use that locally for your tests, but that's not quite the right thing to do. First, you'd need to handle more than dlopen (see CustomElf::GetSymbolPtrInDeps), and to handle them for RTLD_DEFAULT and RTLD_NEXT only.

Comment 3

[BugBot [:suhaib / :marco/ :calixte]]

The bug assignee didn't login in Bugzilla in the last 7 months.
:glandium, could you have a look please?
For more information, please visit auto_nag documentation.

Comment 4

[Mike Hommey [:glandium]]

Since Android 6, we just fallback to the system functions, so it doesn't matter anymore. I don't think we have any interest in making apitrace work on Android < 6.