Re: [PATCH 3/5] pack-objects: clamp negative window size to 0

["Ævar Arnfjörð Bjarmason" <avarab@gmail.com>]

On Tue, May 04 2021, René Scharfe wrote:

> Am 03.05.21 um 16:55 schrieb Jeff King:
>> On Mon, May 03, 2021 at 08:10:24AM -0400, Derrick Stolee wrote:
>>
>>> On 5/1/2021 10:03 AM, Jeff King wrote:
>>>> A negative window size makes no sense, and the code in find_deltas() is
>>>> not prepared to handle it. If you pass "-1", for example, we end up
>>>> generate a 0-length array of "struct unpacked", but our loop assumes it
>>>> has at least one entry in it (and we end up reading garbage memory).
>>>>
>>>> We could complain to the user about this, but it's more forgiving to
>>>> just clamp it to 0, which means "do not find any deltas at all". The
>>>> 0-case is already tested earlier in the script, so we'll make sure this
>>>> does the same thing.
>>>
>>> This seems like a reasonable approach. It takes existing "undefined"
>>> behavior and turns it into well-understood, "defined" behavior.
>>
>> I was on the fence on doing that, or just:
>>
>>   if (window < 0)
>> 	die("sorry dude, negative windows are nonsense");
>>
>> So if anybody has a strong preference, I could be easily persuaded. Part
>> of what led me to being forgiving was that we similarly clamp too-large
>> depth values (with a warning; I didn't think it was really necessary
>> here, though).
>
> There's another option: Mapping -1 or all negative values to the
> maximum:
>
> 	if (window < 0)
> 		window = INT_MAX;
> 	if (depth < 0)
> 		depth = (1 << OE_DEPTH_BITS) - 1;
>
> That's allows saying "gimme all you got" without knowing or being
> willing to type out the exact maximum value, which may change between
> versions.  Not all that useful for --window, I guess.  That convention
> has been used elsewhere I'm sure, but can't point out a concrete
> example.  $arr[-1] get the last item of the array $arr in PowerShell,
> though, which is kind of similar.
>
> Sure, you get the same effect in both cases by typing 2147483647, but
> -1 is more convenient.
>
> Not a strong preference, but I thought it was at least worth
> mentioning that particular bike shed color. :)

That seems sensible to expose, but I think should really be
--window=max, not --window=-1. The latter feels way too much like
assuming that a user might know about C's "set all bits" semantics.

The one example of such a variable I could think of is core.abbrev=no,
which could arguably benefit from a core.abbrev=max synonym.

Another one is *.threads, e.g. grep.threads, index.threads. We currently
say that "auto" is like "max, but I can see how we'd eventually benefit
from splitting those up. I sometimes run git on machines where that
"auto" is 48 or whatever (I haven't benchmarked, but that's surely
counter-productive). Having max != auto in that case (but still having a
"max") would be nice.