- User Since
- Oct 20 2017, 4:34 PM (169 w, 1 d)
Wed, Jan 13
Tue, Jan 12
This goes right before codegen in ThinLTOCodeGenerator. TM.addPassesToEmitFile is the entry point to the llvm backend. So the legacy implementation has this pass ran immediately after the front-end is finished and right before the backend starts. I imagine doing the same here makes the most sense.
Thu, Jan 7
Great, that does indeed seem to work properly. Thanks, Jonas!
I guess if the intention of that is maintain the debugger instance around it should work, but at the moment it segfaults pretty quick with Xcode's lldb using:
On a similar note, I noticed elsewhere after fixing this that you mentioned in a commit message that you intended to remove iplist and ilist and rename as owning_ilist (or something of that sorts). Is that still on the table? Either way, the docs should be updated all around to account for either path name as there are still references to the pre-2016 implementation (e.g. https://llvm.org/docs/ProgrammersManual.html#dss-iplist).
What's the boundaries of the stable API, then? This was a public API that was removed and broke a plugin I used for vim (and I'll be this isn't the only case, just I'm maybe the only one who has worked on lldb before whose tool broke). The author used threading.Thread(someFunc, (debugger,)) to listen on a socket for fetch requests from lldb outside of the prompt. Not the most beautiful of implementations, but it worked for years on top of a promised public stable API.
Wed, Dec 30
Tue, Dec 29
did it backwards
Tue, Dec 22
Hey @ahatanak, I've ran into this problem in my companies projects and was wondering what the status of this patch is?
Dec 9 2020
Yup, this worked. Thank you, Louis! I can also verify that the test suite passed on centOS, Darwin and Windows with a pretty standard build.
Sorry, this was landed already but I forgot to put the Differential Revision: in! Closing it now. As far as the Windows changes you guys are talking about I'll leave up to you -- I'm not a Windows dev and was just maintaining a windows CI oncall last week.
Dec 4 2020
Dec 3 2020
Seems pretty trivially that this just requires LIBCXX_INCLUDE_TESTS and LIBCXX_ENABLE_SHARED for Windows. The error is:
Dec 2 2020
Nov 30 2020
Nov 20 2020
Nov 6 2020
Nov 5 2020
Oct 3 2020
Oct 2 2020
Clean clang-tidy warnings before landing
Oct 1 2020
Update with John's suggestions
Sep 24 2020
Fixed and added a test under the REDUNDANCY prefix.
Fix duplicate inheritance issue
Sep 23 2020
Sep 16 2020
Hmm, I thought we actually just generated a bogus definition for the protocol when it was forward-declared; really, this is better behavior that I expected. Regardless, I don't think it's worthwhile to diagnose this more strongly than a warning because of the history of not doing so.
Sep 8 2020
I don't think it'll actually error out at link time: protocol objects get emitted eagerly on use, cross-module linking is just a code-size optimization. This actually has caused longstanding problems.
A concern that has come up while rewriting this for the listed concerns is forward declared protocols that are defined as non_runtime.
Sep 2 2020
Other than the one comment LGTM.
Aug 28 2020
Looks like it, I did this around the same time the patch landed. Thanks!
Aug 20 2020
Aug 19 2020
This change provides a codegen options flag to clang -fobjc-export-direct-method-wrappers to generate the wrapper functions that begin with the prefix objc_direct_wrapper and are marked as attribute__((alwaysinline)). This way within a link unit the wrapper functions should be inlined away at their call sites, but across a dylib boundary the wrapper call is used.
Aug 11 2020
Adding jmolloy, he seems to be the original author.
Aug 4 2020
No problem! Thank you, John!
Aug 3 2020
Sounds good, just sent out a message to the mailing list.
ping @rjmccall. Any update on a timeline for this review process? Thanks!
Jul 21 2020
Apr 30 2020
Sorry for bump to this old diff, but I agree with both Jim and Greg -- we shouldn't be importing your ~/.lldbinit, but tests shouldn't depend on there never being another br s. This change should land as breakpoint set.
@rjmccall Hey John, I sent the proposal to the addresses I was pointed to but haven't heard back in multiple weeks. Any update on this?
Apr 14 2020
Apr 11 2020
If someone writes up a short proposal for this, with motivation and impact, we'd be happy to present it internally.
Apr 1 2020
Mar 30 2020
Mar 25 2020
Rename and address some issues
Mar 7 2020
Adding some more knowledgeable reviewers for comments on your RFC. I pointed out a few minor nits, but I'll hold off on a technical review until the ObjC-specific details are worked out and there is buy-in on the feature.
Mar 3 2020
Jan 8 2020
Dec 18 2019
This was landed.
Great point! Fixed.
fixup according to Greg's requests
Dec 17 2019
Dec 9 2019
Dec 5 2019
LGTM. Any concerns @clayborg?
Sep 4 2019
Aug 27 2019
Aug 23 2019
Jul 23 2019
Not directly related, but there's some other legacy cruft I figure are valid for removal. e.g.
utils/sync-source misc/grep-svn-log.py git-svn/convert.py