Patch to gcc_update to include revision number in LAST_UPDATED

[Giovanni Bajo]

Mark Mitchell <mark@codesourcery.com> wrote:

>> But there is one important property which svnversion gives you,
>> while your solution does not: whether the working copy is modified
>> or not.
>
> I'm not convinced this is a worthwhile tradeoff.
>
> Yes, it's nice to know that the working copy was modified when looking
> at testresults.  But, testresults can be incorrect for lots of
> reasons:
> bad host compilers, wrong build procedures, etc.  So, I'm not sure
> that getting the "M" on the end of the version is worth the relatively
> significant cost, as shown below.  I think 30 seconds to get the "M"
> seems relatively expensive.


svnversion appears to be very slow in fact. I'm not sure why: the following is
sufficient to tell whether we need the "M" or not:

[ `svn stat -q | sed '/^$/,$d' | wc -l` = "0"] || echo "M"

Or something (i'm no shell freak). I'm confident we can still put that 'M' in
less than 30 seconds.

Giovanni Bajo