Bug 11261 - ccache using wrong cached objects
Summary: ccache using wrong cached objects
Status: RESOLVED WORKSFORME
Alias: None
Product: buildroot
Classification: Unclassified
Component: Other (show other bugs)
Version: unspecified
Hardware: PC Linux
: P5 critical
Target Milestone: ---
Assignee: Arnout Vandecappelle
URL:
Keywords:
Depends on:
Blocks:
 
Reported: 2018-08-29 13:14 UTC by David De Grave
Modified: 2018-09-03 22:47 UTC (History)
2 users (show)

See Also:
Host:
Target:
Build:

Attachments
Add an attachment (proposed patch, testcase, etc.)

Note You need to log in before you can comment on or make changes to this bug.
Description David De Grave 2018-08-29 13:14:45 UTC 

    


    


      



        
          Comment 1
        

        
          Thomas Petazzoni

        

        
        

        
          2018-08-29 20:10:24 UTC
        

      






There is no description in this bug. Something went wrong when submitting the bug ?

    


    


      



        
          Comment 2
        

        
          David De Grave

        

        
        

        
          2018-08-29 23:05:21 UTC
        

      






(In reply to Thomas Petazzoni from comment #1)
Hi Thomas, no, I discovered a problem while building the linux kernel and finally found with the help of Arnout from where came the problem. It's something with ccache and the way buildroot deal with it.

So, he asked me to fill an empty bugzilla report for him to remember (I guess). He will fills all the details himself later... Sorry.

    


    


      



        
          Comment 3
        

        
          Arnout Vandecappelle

        

        
        

        
          2018-09-03 22:47:56 UTC
        

      






Unfortunately, I was not able to reproduce the issue. We wiped the ccache and that solved the problem. I thought that I knew the cause (and thus would be able to reproduce), but apparently I didn't...

The issue was detected while David investigated the libfdt.h issue that was ultimately fixed by replacing -I$(HOST_DIR)/include with -isystem $(HOST_DIR)/include. Before that, he had developed a patch for the kernel that would fix it by changing the libfdt.h in the kernel to match the one in dtc itself. On one build machine, that patch worked perfectly fine; on another one, it apparently had no effect - until the ccache was removed.

In retrospect, it may be possible that this was caused by ccache installed on the host (so not our own ccache). But even with that added in the mix I couldn't reproduce.

I also thought at the time that I understood what was going wrong (I thought it was due to CCACHE_BASEDIR), but I've investigated that option further and it was a dead end.

So, with nothing more to go on, I'm closing this as WORKSFORME. Pity, because unreliable ccache is really painful.