Viewing raw version of a binary file fails to dowload correctly

[gitblit]

Originally reported on Google Code with ID 489

What steps will reproduce the problem?
1. Push a Word Document (docx) to gitblit
2. Browse to the objects location in the tree in Gitblit
3. Click the objects "raw" link

What is the expected output? What do you see instead?
Expected output: Downloads word document to local filesystem. When opened contains
content from that point in time.

What happens: Word document downloaded to local filesystem. When opened contains no
content.

What version of the product are you using? On what operating system?
v 1.6.0, Win 7 x64

Please provide any additional information below.

This bug also impacts other Microsoft Office filetypes, XPS and PDF documents as well
as picture formats (png).

This bug is inconsistent and occurs to some files in some repos and sometimes does
not impact other files or other repos.

Reported by Spitfire5592 on 2014-08-19 16:51:48

[gitblit]

I'm unable to reproduce a 0-length file.

However, I do see that the raw servlet is stamping the response with cache-control
headers which will probably instruct your browser to return cached copies. That was
an unintended regression which is now fixed on master & develop.

Reported by James.Moger on 2014-09-04 17:14:25

• Status changed: Queued
• Labels added: Milestone-1.6.1

[gitblit]

I've reproduced 0-length downloads.  Those should really be 404s because Gitblit can't
find what  the requested resource.  It would be interesting to know the urls which
produced the issue.  Is this on WAR or GO?  Is Gitblit behind a proxy?  Was the repo
in a subdirectory?

Reported by James.Moger on 2014-09-04 19:46:41

[gitblit]

I pushed a fix to properly return a 404 instead of a 0-length download.  But this doesn't
address how you got mangled URLs in the first place.  Issues 478 & 479 may be related.

Reported by James.Moger on 2014-09-04 20:11:33

[gitblit]

v1.6.1 released

Reported by James.Moger on 2014-10-20 21:36:03

• Status changed: Done