johnkenny54

I played around with this a bit and here are some observations. I'm giving relative numbers of errors because I'm running these tests on Windows, and there are a small...

> Do you remember how long that took to run compared to non-multipass? Very roughly twice as long, but since I'm usually doing other work on the computer while these...

I'm not convinced yet that the im-google mismatches are false positives (though they may be in the end). The tests pass if you run them with `{ "name": "convertPathData", "params":...

OK. What do you think about merging this change along with a change in regression.js: `return mismatched === 0;` becomes `return mismatched

The exclude list is more work, but probably worth it. From what I've found in the small number of problems/plugins I've looked at in detail, there's a ton of cleanup...

With the current main branch code, along with PR #1956 and #1961, I can get down to no pixel mismatches by disabling 7 plugins. I ran this with the default...

> > disable these 7 plugins (except possibly convertToShorts and mergePaths) by default > > I strongly disagree. Semi-lossy compression is a large part of SVGO's file size reductions, and...

> > In its current state (with rounding on by default), mergePaths adds 20,424,490 bytes of compression, and introduces 3,424 pixel mismatches in 169 files. With rounding off, there is...

> I know that there's a ``. Will I still not have to worry about it as long as I stick to the outer ``? This comes up a lot...

It looks like the problem here is that there's code in convertTransform that limits the precision of the decomposition to no more than the maximum number of decimal digits in...