Tags: facebook/fbthrift
Tags
Generate ServiceMetadata from schema.thrift Summary: Similar to previous diff, but for ServiceMetadata. Reviewed By: thedavekwon Differential Revision: D82437949 fbshipit-source-id: 86e561733e650d5bbf72a5084a9a5778ae64f090
Updating submodules Summary: GitHub commits: facebook/buck2-prelude@9615d24 facebook/fb303@2f19bdc 531d08e facebook/folly@af67dd1 facebook/mvfst@1d59def facebook/proxygen@4c92179 facebook/wangle@b3688d2 facebookexperimental/edencommon@bfe92ce facebookexperimental/rust-shed@2a8cad8 facebookexternal/wrist-factory@0f0e99d facebookincubator/fizz@2a347d6 Reviewed By: JurjenLelifeld fbshipit-source-id: 98909c27e30122295cbfab187f2dffca1706f65e
Remove empty explicit URI annotation from typedefs in built-in any, p… …rotocol and Object Summary: Remove explicit `thrift.Uri` annotation from built-in typedefs, in preparation for removing `scope.Typedef` from `thrift.Uri`. Since these typedefs are not annotated with `thrift.AllowLegacyTypedefUri`, the compiler already enforces an empty URI. Reviewed By: thedavekwon Differential Revision: D82867585 fbshipit-source-id: 536debcbe38de1569b5bdd6a6db2acfe874d4e6c
Updating submodules Summary: GitHub commits: facebook/fb303@27a5fcb f347d0f facebook/mvfst@afb9c76 facebook/proxygen@84bf764 facebook/wangle@c5b06b1 facebookexperimental/edencommon@5a892ce facebookexperimental/rust-shed@9a29930 facebookincubator/fizz@8c9e4fe Reviewed By: ajb85 fbshipit-source-id: 875229cec2d10346bef018ad5618646d65ce5ae8
Revert D81597877 Summary: This diff reverts D81597877 T237166478 Depends on D81597877 Reviewed By: Mizuchi Differential Revision: D81914786 fbshipit-source-id: ae96ef595d31caefe1afe1044c928998c8d5d7a1
fbcode/thrift/lib/thrift/ Differential Revision: D81300705 fbshipit-source-id: 943ab492f1948f23ca2a416edea9574851267b6a
Updating submodules Summary: GitHub commits: facebook/fb303@17a2ab8 fd711e6 facebook/folly@17d4416 facebook/mvfst@3db09c5 facebook/proxygen@eddcb6e facebook/wangle@38498f7 facebookexperimental/edencommon@afa1298 facebookexperimental/rust-shed@48c915e facebookincubator/fizz@31c8449 meta-pytorch/torchcomms@a232709 Reviewed By: JurjenLelifeld fbshipit-source-id: 1456c625902cf6e292039534ed62f9090da926bc
basic sink tests Summary: These tests cover all the basic cases w/r/t regular sink-ing behavior, early returns, expected and unexpected exceptions. Reviewed By: prakashgayasen Differential Revision: D79922296 fbshipit-source-id: 873a161d11fdef2b36eabaefdcff26ac700520a8
Decouple serialization and types more
Summary:
We are decoupling serialization and types to speed up modular builds. This is based on the observation that serialization code consists of large templates that few users need to consume directly. Decoupling is mostly complete on the target level and the remaining part is making the `<module>-cpp2-serialization` target a private dependency of `<module>-cpp2-types` as in the following diagram (currently it is a public dependency):
```
[<module>-cpp2-types-headers]
^ ^
(public)| |(public)
| |
[<module>-cpp2-types] --> [<module>-cpp2-serialization]
(private)
```
The main remaining blocker is that the generated `<module>_types.cpp` includes `<module>_types.tcc` which is a header currently combining some implementation details of types and serialization code. It also contains some remnants of serialization code itself even after most of it has been moved to separate translation units.
This diff decouples serialization from types by moving the serialization code that is currently in `<module>_types.tcc` to `<module>_custom_protocol.h` and the one that is currently in `<module>_types.cpp` to `<module>_types_serialization.cpp`. `<module>_custom_protocol.h` headers are arguably a better fit for this than `<module>_types.tcc` which were often misused (D79115339).
This also prevents users from creating semi-broken targets with some headers not includable due to long paths on Windows (see e.g. D79198470).
And last but not least it slightly improves build speed as measured by `benchmark_diff --diff D79106423 --top-slow 1000`:
```
+-----------------------------------------------------------------+------------+--------------+----------------------+--------------------+
| Measurement name | End date | All fbcode | Monetization Users | Monetization Org |
| https://www.internalfb.com/phabricator/paste/view/P1898352612 | | avg EYS | avg EYS | avg EYS |
+=================================================================+============+==============+======================+====================+
| benchmark/D79106423/1754691520_a2f9eb1c | 2025-08-08 | 0.01398 | 0.02013 | 0.01664 |
+-----------------------------------------------------------------+------------+--------------+----------------------+--------------------+
```
Per week measurement data in markdown:
[07/11-07/18](https://www.internalfb.com/phabricator/paste/view/P1898344930) - -0.00068 EYS - 1875 builds - Top user: nikitabhalla: 0.00307 (-451.64%)
[07/18-07/25](https://www.internalfb.com/phabricator/paste/view/P1898350876) - 0.04267 EYS - 1724 builds - Top user: bzan: 0.00317 (7.44%)
[07/25-08/01](https://www.internalfb.com/phabricator/paste/view/P1898348469) - 0.00424 EYS - 1713 builds - Top user: chenyh: 0.00329 (77.74%)
[08/01-08/08](https://www.internalfb.com/phabricator/paste/view/P1898352431) - 0.00969 EYS - 2058 builds - Top user: neelansh: 0.00511 (52.72%)
More context: T206890027.
Reviewed By: iahs
Differential Revision: D79106423
fbshipit-source-id: 54897209026837371f5ceadf6d9ab0f41d466cf8
fix blame on unexpected sink exception Summary: Small re-factor so that if the sinks throws an unexpected exception on the sink side, the blame gets encoded properly. If we don't like that unexpected exception is wrapped in `ApplicationError`, we can do a bigger refactor so sink has its own custom handling of unexpected exception. Reviewed By: prakashgayasen Differential Revision: D79474562 fbshipit-source-id: e4e2b48f7aa92848331e96e38880d0943fc9a41e
PreviousNext