Walking children commits?

Walking children commits?

[Mike Hommey]

Hi,

I am trying to get all the children commits of a set of commits. To do
that, I'm using this:

  git rev-list --topo-order --ancestry-path --boundary <all heads> ^A ^B ...

where A, B, and following are those commits I want the children of. From
the gitrevisions documentation, it seems that should get me what I want,
but if I build a dag manually (from the output of rev-list --parents)
and get all the children of those commits, I get more results than what
the command above returns (and that command also gives false positives).

Am I doing something wrong, or am I hitting a bug in --ancestry-path
walking? In the latter case, I'll try to find a reduced test case,
because currently, this involves a 260k commit, 1.5GB repository with
close to 2k heads and a set of 153 commits to find the children of.

Cheers,

Mike

Re: Walking children commits?

[Mike Hommey]

On Tue, Mar 31, 2015 at 08:02:09AM +0900, Mike Hommey wrote:
> Hi,
> 
> I am trying to get all the children commits of a set of commits. To do
> that, I'm using this:
> 
>   git rev-list --topo-order --ancestry-path --boundary <all heads> ^A ^B ...
> 
> where A, B, and following are those commits I want the children of. From
> the gitrevisions documentation, it seems that should get me what I want,
> but if I build a dag manually (from the output of rev-list --parents)
> and get all the children of those commits, I get more results than what
> the command above returns (and that command also gives false positives).

The false positives are actually boundaries other than A, B, etc. so I'd
actually need to remove --boundary and add A, B to the result manually.
That still leaves the missing results, though.

Mike

Re: Walking children commits?

[Mike Hommey]

On Tue, Mar 31, 2015 at 08:35:52AM +0900, Mike Hommey wrote:
> On Tue, Mar 31, 2015 at 08:02:09AM +0900, Mike Hommey wrote:
> > Hi,
> > 
> > I am trying to get all the children commits of a set of commits. To do
> > that, I'm using this:
> > 
> >   git rev-list --topo-order --ancestry-path --boundary <all heads> ^A ^B ...
> > 
> > where A, B, and following are those commits I want the children of. From
> > the gitrevisions documentation, it seems that should get me what I want,
> > but if I build a dag manually (from the output of rev-list --parents)
> > and get all the children of those commits, I get more results than what
> > the command above returns (and that command also gives false positives).
> 
> The false positives are actually boundaries other than A, B, etc. so I'd
> actually need to remove --boundary and add A, B to the result manually.
> That still leaves the missing results, though.

Ah, the missing results are commits that both are part of the set of
commits and are heads...
So all in all --ancestry-path works for my use case, but I need
post-processing.

Mike