On 9/6/19 12:58 AM, Jeff King wrote:
> On Fri, Sep 06, 2019 at 12:53:49AM +0200, Stephan Beyer wrote:
>
>> On 9/6/19 12:48 AM, Jeff King wrote:
>>> On Thu, Sep 05, 2019 at 10:24:59AM +0200, Stephan Beyer wrote:
>>>
>>>> Compiler heuristics for detection of potentially uninitialized variables
>>>> may change between compiler versions and enabling link-time optimization
>>>> may find new warnings. Indeed, compiling with gcc 9.2.1 and enabled
>>>> link-time optimization feature resulted in a few hits that are fixed by
>>>> this patch in the most naïve way. This allows to compile git using the
>>>> DEVELOPER=1 switch (which sets -Werror) and using the -flto flag.
>>>
>>> Lots of discussion in this thread. Let's try to turn it into some
>>> patches. :)
>>
>> I thought the same and just sent my version of it. :D
>
> Yeah, I see that our mails crossed. :) I like some of my versions
> better, but I'm OK with either (or a mix-and-match).
I took a quick glance at yours. I also noticed the issue you address in
[PATCH 2/6], but I was unsure if this is the way to go (I'm only
occasionally reading on this list). I would prefer your patch series,
with maybe one exception...
The thing is: I had *exactly* the same commit like your [PATCH 6/6]
(except for the commit message and for the number), but I dropped it.
Why? Because I had the feeling (no particular instance though) that the
second locate_object_entry_hash() for each insertion *can* indeed take
"too much" time. Also, I was wondering, if the "found = 1" case should
be catched as a BUG("should not happen") or something.
I don't care much, though. The performance impact should probably be
checked carefully.
Stephan
