Bug with rev-list --reverse?

Hi,

If I do these:

% git log --oneline -1 v1.8.1.5^..v1.8.1.6 % git log --oneline --reverse -1 v1.8.1.5^..v1.8.1.6

I expect to get a different output, and not both showing v1.8.1.6. Wouldn't you agree?

Cheers.

On Thu, Apr 18, 2013 at 05:17:14AM -0500, Felipe Contreras wrote:

I expect to get the same output. This is probably because I consider --reverse to be an output filter. So I expect to show the commits "v1.8.1.5^..v1.8.1.6 -1" which selects a single commit and then show that in reverse order.

On Thu, Apr 18, 2013 at 5:26 AM, John Keeping <john@keeping.me.uk> wrote:

How about this:

% git log --oneline --reverse --max-count=1 v1.8.1.5^..v1.8.1.6

In this case --max-count is acting as "start from the first commit before the tip", not as "output a maximum of one commit". Given that the name is max-count, I expect it to be the later.

And if max-count doesn't select a maximum of n commits, then what does?

Felipe Contreras:

  Quoting the manual page:

  Commit Limiting
    Besides specifying a range of commits that should be listed using the special notations explained in the description, additional commit limiting may be applied. Note that they are applied before commit ordering and formatting options, such as --reverse.

Given that, I would expect the output to be the same.

On Thu, Apr 18, 2013 at 5:47 AM, Peter Krefting <peter@softwolves.pp.se> wrote:

If expectations were based on documentation, all one has to do is document bugs, and there would be no bugs anymore :)

Code can be changed to fit more appropriately user expectations (which are independent of documentation), and the documentation updated accordingly.

Felipe Contreras <felipe.contreras@gmail.com> writes:

It's been this way forever, and applies to rev-list where we can't just break how options work (for fear of breaking scripts).

You could come up with a patch series that first starts emitting warnings whenever the user asks for behavior that will change, and later flips the default and removes the warning (the latter would be merged for 2.0 or so).

Thomas Rast <trast@inf.ethz.ch> writes:

Please don't. The fact that "reverse then count" mode may be useful does not mean "count then show in reverse" mode does not have any use. "git log --oneline --reverse -20" is a very valid way to ask "what did I do recently" and get "you did this, that and then..." in that order.

Adding a new option can be done anytime without any complex transition plan. You may want to introduce a --show-in-reverse synonym to the current --reverse when you add the new mode of reversing (--reverse-before-count?) so that eventually we won't have to ask "which kind of reverse an unadorned --reverse option mean?" by deprecating a plain "--reverse", though.