max_preview_size isn't active for changeset viewing

[Dan <trac@…>]

I'm not sure if this is a defect, or if it's intended, but I have a very large text file (it's test input data) in my repository. When I attempt to view it in the browser, I correctly get a message that the file size is too large to preview. However, when I try to view a changeset regarding this file, Apache just goes away to contemplate its navel, and meanwhile access to the site is non-existent. I have never waited long enough to view the final results.

Maybe max_preview_size should control the changeset view as well, or maybe there should be another variable. I realize that the "difference" may be much smaller than the total file, but in this case, it is causing a lot of grief.

I think that a work-around might be to change the mime-type of the file in the repository, but I haven't tried that yet. Maybe then Trac won't try to do anything with the file in the changeset view.

If it matters, I am using Apache 2.0.52, Python 2.4.2 with mod_python, under Windows x64.

[anonymous]

This is not only a problem for apache spending a lot of time sending out the changeset page but also for the browser to actually display the page. I've crashed firefox a few times (the process was using all available memory trying to display the big HTML page).

This can also arise in case of important changes to a lot of different files. In that case, a paging solution could be helpful, having a next button to see the next diff.

[Christian Boos]

See also #2591.

[Christian Boos]

The configuration option

[changeset]
max_diff_bytes = 20000000

added in r2968 can be used for that.

Note that the above number is a limit for the sum of the size of all old and new files involved in a changeset.

Therefore, a limit of 20Mb actually correspond to the possibility of showing diffs for, e.g. 10 modified files of 1Mb each.

Note also that when requesting a diff precisely on a very big file (e.g. by the use of the Last Change link), the size limit will not be taken into account and the diffs will always be computed.