-
Notifications
You must be signed in to change notification settings - Fork 52
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
FTBFS on AArch64 Darwin due to failed alignment assertion #831
Comments
rodaine
pushed a commit
to bufbuild/protovalidate-cc
that referenced
this issue
Aug 7, 2024
Updates Protobuf to v27 and protovalidate to v0.7.1, and fixes all of the resulting compilation and conformance failures. As one would expect, there was a tremendous amount of troubleshooting involved in this thankfully-relatively-small PR. Here's my log of what happened. I'll try to be succinct, but I want to capture all of the details so my reasoning can be understood in the future. - First, I tried to update protobuf. This led to pulling a newer version of absl. The version of cel-cpp we use did not compile with this version of absl. - Next, I tried to update cel-cpp. However, the latest version of cel-cpp is broken on macOS for two separate reasons <sup>[1](google/cel-cpp#831), [2](https://github.com/google/cel-cpp/issues/832)</sup>. - After taking a break to work on other protovalidate implementations I returned and tried another approach. This time, instead of updating cel-cpp, I just patched it to work with newer absl. Thankfully, this proved surprisingly viable. The `cel_cpp.patch` file now contains this fix too. - Unfortunately, compilation was broken in CI on a non-sense compiler error: ``` error: could not convert template argument 'ptr' from 'const google::protobuf::Struct& (* const)()' to 'const google::protobuf::Struct& (* const)()' ``` It seemed likely to be a compiler issue, thus I was stalled again. - For some reason it finally occurred to me that I probably should just simply update the compiler. In a stroke of accidental rubber-ducking luck, I noticed that GitHub's `ubuntu-latest` had yet to actually move to `ubuntu-24.04`, which has a vastly more up-to-date C++ toolchain than the older `ubuntu-22.04`. This immediately fixed the problem. - E-mail validation is hard. In other languages we fall back on standard library functionality, but C++ puts us at a hard impasse; the C++ standard library hardly concerns itself with application-level functionality like SMTP standards. Anyway, I channeled my frustration at the lack of a consistent validation scheme for e-mail, which culminated into bufbuild/protovalidate#236. For the new failing test cases, we needed to improve the validation of localpart in C++. Lacking any specific reference point, I decided it would be acceptable if the C++ version started adopting ideas from WHATWG HTML email validation. It doesn't move the `localpart` validation to _entirely_ work like WHATWG HTML email validation, as our version still has our specific checks, but now we are a strict subset in protovalidate-cc, so we can remove our additional checks later if we can greenlight adopting the WHATWG HTML standard. - The remaining test failures are all related to ignoring validation rules and presence. The following changes were made: - The algorithm for ignoring empty fields is improved to match the specified behavior closer. - The `ignore` option is now taken into account in addition to the legacy `skipped` and `ignore_empty` options. - Support is added for `IGNORE_IF_DEFAULT_VALUE` - An edge case is added to ignore field presence on synthetic `Map` types. I haven't traced down why, but `has_presence` seems to always be true for fields of synthetic `Map` types in the C++ implementation. (Except in proto3?) And with that I think we will have working Editions support.
@jcking Any thoughts on this issue? I think this breaks macos toolchain. |
I believe I removed this assertion last week. Removing it should be safe. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Unfortunately, it seems that compatibility with some ABIs was probably broken in abbbf51.
The reasoning for this assertion to fail is explained well in this Stack Overflow answer, though currently it is unclear to me whether the code actually is broken. I'm not sure why this assertion failing would cause any problems.
The text was updated successfully, but these errors were encountered: