[PATCH] Really skip abi checks when requested

Amit Kucheria amit at canonical.com
Thu Jan 22 14:44:53 GMT 2009


On Thu, Jan 22, 2009 at 4:33 PM, Tim Gardner <tim.gardner at canonical.com> wrote:
> Amit Kucheria wrote:
>> On Thu, Jan 22, 2009 at 3:58 PM, Tim Gardner <tim.gardner at canonical.com> wrote:
>>> Amit Kucheria wrote:
>>>> Hi,
>>>>
>>>> Currently, when you wish to ignore ABI, you still need to provide the
>>>> old and new ABI files else the build fails. Failure of 2.6.28-5.13 is
>>>> due to this reason, although Tim added the ignore files to -5.12.
>>>>
>>>> Ben, could you explain (once again) why providing the ABI is
>>>> desirable?  I am probably too dense to understand the reasoning for
>>>> this.
>>>>
>>>> Attached patch attempts to fix it my way.
>>>>
>>>> /Amit
>>>>
>>> How about this minor correction.
>> +
>> +     # Exit if the ABI files are missing, but return status based on whether
>> +     # skip ABI was indicated.
>> +     if ("$abiskip" eq "1") {
>> +             exit(0);
>> +     } else {
>> +             exit(1);
>> +     }
>>  }
>>
>> Isn't the point that we _should not_ exit when we have requested
>> abiskip? Your correction functionally reverts it back to what it is.
>>
>> /Amit
>>
>
> At this point in the script you've determined that the previous ABI
> files don't exists, so you cannot proceed. However, you've requested to
> skip ABI checking which implies the script should return success
> regardless. Otherwise the build fails on an error status.

Duh. I forgot that the return status caused the build to fail, not the
actual exit() call.

Ack.

/Amit



More information about the kernel-team mailing list