pitrou commented on PR #40508:
URL: https://github.com/apache/arrow/pull/40508#issuecomment-2025145409

   So, I'm able to get some insight by using the `strings` command followed by 
some processing:
   ```console
   $ strings wheel-3.8-successful/pyarrow/libarrow.1600.dylib | \grep "^_\w" | 
sort | uniq | sed -E 's/^_+/_/g' | c++filt > strings-successful.txt 
   $ strings wheel-3.8-failed/pyarrow/libarrow.1600.dylib | \grep "^_\w" | sort 
| uniq | sed -E 's/^_+/_/g' | c++filt > strings-failed.txt 
   ```
   
   Then I can diff `strings-failed.txt` and `strings-successful.txt` to get an 
idea of the differences:
   ```console
   $ diff --color -u strings-failed.txt strings-successful.txt
   ```
   
   And amongst the changes some have to do with the protobuf-generated code for 
ORC:
   ```diff
   @@ -60127,232 +59930,265 @@
    orc::proto::Encryption::GetMetadata() const
    orc::proto::Encryption::ByteSizeLong() const
    orc::proto::Encryption::GetClassData() const
   +orc::proto::Encryption::GetCachedSize() const
    orc::proto::Encryption::IsInitialized() const
   -orc::proto::Encryption::AccessCachedSize() const
   +orc::proto::Encryption::SetCachedSize(int) const
   ```
   
   (note: "-" lines are the failed wheel build, "+" lines are the successful 
wheel build)
   
   This indicates that between the failed and successful build, two different 
versions of the protobuf compiler were invoked:
   - the failed build uses a protobuf compiler that contains 
https://github.com/protocolbuffers/protobuf/commit/ecd106bd49b700b862f1a1241800988bb8ade120
 (" Replace GetCachedSize+SetCachedSize with a single AccessCachedSize 
function"), therefore is v26.1 or later
   - the successful build uses a previous protobuf compiler, therefore v26.0 or 
earlier.
   
   Now to understand where these two different protobuf versions come from... 
@kou @assignUser  any idea?


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]

Reply via email to