Move Lib already loaded check before dlopen Call

[thestr4ng3r]

for rizinorg/rz-ghidra#38

This prevents constructors, for example doing C++ static initialization, from being executed twice.

[radare]

My concern on this code was mainly to lod the same plugin twice just because its copied in multiple paths or its the same with different name. Not sure if we should use checksums add another check to avoid this to happen

…

On 19 Sep 2019, at 13:02, Florian Märkl ***@***.***> wrote: for rizinorg/rz-ghidra#38 This prevents constructors, for example doing C++ static initialization, from being executed twice. You can view, comment on, or merge this pull request online at: #15075 Commit Summary Move Lib already loaded check before dlopen Call File Changes M libr/util/lib.c (38) Patch Links: https://github.com/radareorg/radare2/pull/15075.patch https://github.com/radareorg/radare2/pull/15075.diff — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub, or mute the thread.

[thestr4ng3r]

Checksums break again if you have a slightly different version of the same plugin. One good way would be to add a unique id string per plugin to RLibStruct and check that after loading the lib and before registering the plugin. But we still need to make sure not to call any constructors before we haven't checked the duplication. I don't know enough about dynamic loading to say if this is possible on all platforms.

[radare]

yes, thats what we must check.. but if c++/objc crap run things when you do dlopen.. not sure how can we handle this in the right way without using RBin instead

[radare]

lets merge for now, most people delete all the plugins after every r2 update, as well as most people put them all in the same place. so it shouldnt be a problem

[thestr4ng3r]

Yes, this patch only makes the situation better. Actually even after it there is a special case when you have a duplicate plugin and the one that is loaded first triggers the Module version mismatch /home/florian/bin/prefix/radare2/lib/radare2/core_ghidra.so (3.8.0) vs (4.0.0-git), then it still breaks. I have to investigate whether this is fixable somehow.