pitrou opened a new issue, #35641:
URL: https://github.com/apache/arrow/issues/35641

   ### Describe the bug, including details regarding any error messages, 
version, and platform.
   
   Example here on the AMD64 Conda C++ build:
   
https://github.com/apache/arrow/actions/runs/5005430904/jobs/8969381538?pr=33634
   
   ```
   === ccache statistics before build
   ===
   Cacheable calls:   284 / 894 (31.77%)
     Hits:             14 / 284 ( 4.93%)
       Direct:         14 /  14 (100.0%)
       Preprocessed:    0 /  14 ( 0.00%)
     Misses:          270 / 284 (95.07%)
   Uncacheable calls: 610 / 894 (68.23%)
   Local storage:
     Cache size (GB): 0.1 / 1.0 (11.35%)
     Hits:             14 / 284 ( 4.93%)
     Misses:          270 / 284 (95.07%)
   
   [ ... CMake runs ... ]
   
   real 27m46.563s
   user 49m43.936s
   sys  4m28.705s
   
   ===
   === ccache statistics after build
   ===
   Cacheable calls:    568 / 1788 (31.77%)
     Hits:             298 /  568 (52.46%)
       Direct:         233 /  298 (78.19%)
       Preprocessed:    65 /  298 (21.81%)
     Misses:           270 /  568 (47.54%)
   Uncacheable calls: 1220 / 1788 (68.23%)
   Local storage:
     Cache size (GB):  0.1 /  1.0 (11.35%)
     Hits:             298 /  568 (52.46%)
     Misses:           270 /  568 (47.54%)
   ```
   
   So ccache was completely successful in this build (0 misses during the 
entire CMake configure / build), yet CMake took 27 minutes.
   
   We should investigate:
   * disabling building static libraries (`ARROW_BUILD_STATIC`) to save on 
linking times
   * why there are more than 600 uncacheable calls and also 280 calls, which is 
the majority of the ~860 compilation calls (the "Building CXX" lines in the 
logs)
   
   Note that the "AMD64 Conda C++" build can be reproduced using 
[Archery](https://arrow.apache.org/docs/dev/developers/continuous_integration/docker.html)
 with the `conda-cpp` image:
   ```console
   $ archery docker run conda-cpp
   ```
   
   
   ### Component(s)
   
   C++


-- 
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