Re: [PATCH] perf: Fix endianness argument compatibility

[David Miller]

From: "Ian Munsie" <imunsie@xxxxxxxxxxx>
Date: Tue, 13 Apr 2010 18:37:33 +1000

> From: Ian Munsie <imunsie@xxxxxxxxxx>
> 
> Parsing an option from the command line with OPT_BOOLEAN on a bool data
> type would not work on a big-endian machine due to the manner in which
> the boolean was being cast into an int and incremented. For example,
> running 'perf probe --list' on a PowerPC machine would fail to properly
> set the list_events bool and would therefore print out the usage
> information and terminate.
> 
> This patch makes OPT_BOOLEAN work as expected with a bool datatype. For
> cases where the original OPT_BOOLEAN was intentionally being used to
> increment an int each time it was passed in on the command line, this
> patch introduces OPT_INCR with the old behaviour of OPT_BOOLEAN (the
> verbose variable is currently the only such example of this).
> 
> I have reviewed every use of OPT_BOOLEAN to verify that a true C99 bool
> was passed. Where integers were used, I verified that they were only
> being used for boolean logic and changed them to bools to ensure that
> they would not be mistakenly used as ints. The major exception was the
> verbose variable which now uses OPT_INCR instead of OPT_BOOLEAN.
> 
> Signed-off-by: Ian Munsie <imunsie@xxxxxxxxxx>

Thanks for finding and fixing this bug.
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/