Am 16.07.2005 um 07:58 schrieb Michael Rueger:
Doug Way wrote:
And Ken is probably right that perhaps we should bump up the version number. 3.8.1? Hmm, well certainly we were going to do that if we added any updates, although in this case it's simply a fixed image with the same update number (and same source code), so it's a close call.
I would vote against a version number change, but rather change the name of the download file or something? Or, we could release an update to fix the proxy problem, so the image will be 3.8 6666. There is already a pending update (the positionOf... fix) plus possibly Goran's FatSocketStream.
And this could make sense:
http://bugs.impara.de/view.php?id=1382
0001382: 3.8 performance regression - ByteSymbol>>#hash Description Magma's peak commit rate in 3.8 dropped from 24 per second in 3.7 to 18 per second in 3.8. I traced the cause of this problem to ByteSymbol>>#hash. Because of the new String hierarchy to support Multi-byte Strings, it is no longer benefitting from the primitive for calculating its hash that used to be in String>>#stringHash:initialHash:. It was moved to ByteString but not to ByteSymbol so now, ByteSymbol is inheriting the slow one from String which, per the comment, works for WideStrings too.
This fix simply restores the method with the primitive call that regular [Byte]Symbols have always benefitted from.
I added this to 3.9a already, and I use it it my work-image for a week now.
Marcus