Thomas Schuerger wrote:
> The (wrong) range check has been in gtkg since revision 1.33 of
> core/downloads.c from 08-Jan-05.

Heh? That's about 3 years in the future from now.

> This means that the release versions 0.95.1 to 0.95.3 are affected
> by the bug.

Whereas 0.95 would simply crash if fed a inconsistent range - just
in case anyone plans to downgrade because of this.
 
> It seems that with these versions, downloads can succeed only if
> servents which don't return "Content-Range" headers are involved,
> because in that case, the range intersection check is skipped.

Correct me if I'm wrong but I think this matters only if the last byte
of a file has become its own range. Obviously that happens sometimes
but I haven't noticed it in 5 month.

> Sounds like a new release is necessary soon...

Didn't you say that you've seen the "from < to" assertion triggering?
There's also still the bug that seems to cause lossage of downloaded
chunks. I'd say a new version shouldn't be release right now, if
this/those bugs can be fixed within the next few days because those
look more serious than this one to me.

-- 
Christian

Attachment: pgpGVOzWq1wUL.pgp
Description: PGP signature

Reply via email to