On 02/12/2014 02:34 PM, Adrien Destugues wrote:
Thanks for following up on the suggestions. Personally, I think "result"
is still not a good name, I prefer "written" (or "read", depending on
context), but this is entering the area of nitpicking.
I went with result, because in this case the variable isn't actually
used as a count/size: we only check the "negative error code" case.
The more I think about this whole change, the more I think it sucks (at
least as it is now). You switched from an interface with a Write()
method that either writes everything or fails (BDataOutput) to an
interface with a Write() method that explicitly supports partial writes
(BDataIO). I was wondering why the incorrect implementation of
PackageFile::IORequestOutput::Write() in packagefs (for which Jérôme
added a TODO) doesn't break things completely. But obviously that's
because the code using the interface hasn't been adjusted to handle the
return value correctly either.
I think it would be better to revert the interface switch and rather
introduce a BDataOutput -> BDataIO adapter. This would allow using
BDataIOs with ZlibDecompressor while keeping the more concise interface.
