Re: sconsign database discussion

On 2/23/21 8:59 AM, anatoly techtonik wrote:
> On Tue, Feb 23, 2021 at 6:02 PM Mats Wichmann <[email protected]> wrote:
>>
>> I did get a quick implementation going, and the informal timing I did on
>> that were somewhat disappointing.
> 
> Disappointing for SQLite or for pickled dict?

disappointing for the use of SQLite.

> Is there still an easy way to do SCons benchmarks and share them?
> 

there's still code in bench/ and in timings/.  It's rotted a bit... I've 
done sporadic updates for Python 3 where it was obvious, but some of the 
tests are just not useful - e.g/ there's one that tries to time scanning 
for includes, and tries to scale itself to an expected amount of time - 
and the alogirithm generates so many includes for today's processor 
speeds that no compiler will accept them (Bill noted that that test was 
scaled for a Celeron-class processor).