On 03/19/2012 01:06 PM, Henrik Holst wrote: > Considering that the failing file in question is 3.5GiB it's probably a > signed 32-bit problem with the size and/or range in either wget or the > server. Would be interesting to see the range requests done by your version > of wget around tje.signed 32 limit.
A very pertinent point, that I'm surprised hadn't occurred to anyone else yet. Wget has had >2GB support since at least 1.10.x, but I think it may need to be built in - it's entirely possible that copy of wget was built without it (somehow). It's also conceivable that 32-bit file problems persist in the Windows code, that is fixed in the Unix code. Or it could be problems with the system libraries for Cygwin or MinGW (whichever was used to build it). And it could also be the server failing to handle ranged requests reaching into that domain, though that would be a strange problem indeed for kernel.org to have... -mjc
