HTTP content-range

[PerilousApricot]

Hello,

When accessing files over HTTPS, accessing subsets of the file with the Range header returns the following header for a 16kB read content-read: bytes 0-16384/16385.

According to the HTTP spec, however, the number after the slash (e.g. 16385) should be the length of the whole document, and not the length of the range that was just returned (ref [1] [2]).

I can trigger this on 4.8.5-1.osg34.el7

Cheers
Andrew

[1] https://tools.ietf.org/html/rfc7233#page-12
[2] https://developer.mozilla.org/en-US/docs/Web/HTTP/Headers/Content-Range

[PerilousApricot]

BTW, Content-Length is correctly set to 16385, so I can't use that instead to work around the glitch

[bbockelm]

The report looks correct to me.

To help us determine severity - how did you notice this problem? Is there a particular client that chokes on the incorrect response?

[PerilousApricot]

Looks like this should be filesize instead of cnt on this line:

xrootd/src/XrdHttp/XrdHttpReq.cc

Line 1928 in 7407ea8

sprintf(buf, "%lld-%lld/%d", rwOps[0].bytestart, rwOps[0].byteend, cnt);

I'm working on a pure-python backend for uproot based on the HTTP api (so it doesn't have to have the xrootd C++ API), so there's no users yet. As a workaround, I can start with a HEAD call to get the file length.

[bbockelm]

Could you send in a PR for this? Would be nice to squeeze this in prior to 4.9.0.

[PerilousApricot]

Eyeballing it, this looks like a simple change: #876

[abh3]

I believe this was fixed, was it not?

[PerilousApricot]

Have a server with this patch up to test?

On Fri, Apr 19, 2019 at 19:46 Andrew Hanushevsky ***@***.***> wrote: I believe this was fixed, was it not? — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#875 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AAAWZKUGJ33F42INOGWEEGLPRJRWRANCNFSM4GJ2XVNQ> .

-- It's dark in this basement.

[simonmichal]

@PerilousApricot : your patch was included in 4.9.0, can we close this one?

[abh3]

Fabrizzio, has this been fixed as I think it has?

[ffurano]

Well, the fix seems correct to me, and I have not noticed any troubles since its commit