RFR: 1236: Add jcheck option to check for large binary files [v3]

[Erik Joelsson]

On Sat, 4 Dec 2021 15:27:48 GMT, Guoxiong Li <gli at openjdk.org> wrote:

> > I think you need to figure out the size by inspecting the binary patch. I'm not sure exactly how, but it looks like all the binary data is encoded in the patch hunks.
> 
> I dive into the diff information to find the size. Only the added and deleted files have the entire size. The modified files have the delta size (modified size) which we can't use. And the renamed and copied files have no size, they only have the source path and target patch.
> 
> May need help from the experts who are famaliar with Git/HG.

I think size delta in the case of modified is fine. In my experience, it's very rare for binary patches to actually be partial in real situations. When replacing an image or archive, the whole file usually changes enough. If the binary patch is empty, due to file moved/rename, then we can't really do anything about that unfortunately. Have you looked at actual patch output from Git to see what it looks like for renamed binary files? Git doesn't track file renames explicitly, so I'm not sure what would actually happen here.

-------------

PR: https://git.openjdk.java.net/skara/pull/1247