[cfe-dev] The effects of exact diagnostic checking

[Ted Kremenek]

On Feb 7, 2009, at 2:26 PM, Sebastian Redl wrote:

> Mike Stump wrote:
>> :-(  This is kinda annoying.  If 100 people did this with 10 test
>> cases apiece, at any given time, there would be about 1000 failing
>> test cases, and the poor developers would then have to puzzle out
>> which failures are theirs, which failures mark real bugs in their  
>> work
>> that they now can't catch and so on.  Imagine we have a cron job that
>> reverts all work that doesn't result in a clean test suite run.
> Then my change would be reverted :-)
> The problem is that there is no good way of doing such a thing without
> temporarily breaking the tree. If I just published my patch to the
> mailing list and asked people to apply it and see what test failures
> they have and if they can fix them, would it happen? The people have  
> the
> same work as before - puzzle out which failures are there, and which  
> are
> real bugs - plus they have to download and manually apply the patch,
> instead of having SVN do it for them.


Hi Sebastian,

We really do appreciate this work, but I do think giving people heads  
up about this *before* the patch was applied would have been better.   
People could have then played around with the patch, see how it broke  
things, and then have fixes ready soon after the patch went it.  Even  
if the only way to do this was to temporarily break the tests in the  
tree, it probably would be better to do it during the week instead of  
on the weekend (where tests could linger, broken, for several days).

Ted