threads / patch / 15413

patchRevWalk: Fix RevSort.REVERSE combined with RevSort.TOPO

Subject: [JGIT PATCH] RevWalk: Fix RevSort.REVERSE combined with RevSort.TOPO

## tl;dr

4 messages between Sep 7, 2008 and Sep 7, 2008. Diffs are folded; open one to read it.

replies: 3people: 2as markdown or json

Shawn O. Pearce· Sep 7, 2008, 21:00 UTC · lore

From: Shawn O. Pearce <sop@google.com>

The reverse sort was using the initial pending queue, not the generator that comes before it in the workflow. This caused a combination of TOPO and REVERSE to return no commits.

Signed-off-by: Shawn O. Pearce <sop@google.com>
---
 .../org/spearce/jgit/revwalk/StartGenerator.java   |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)

Show changes to org.spearce.jgit/src/org/spearce/jgit/revwalk/StartGenerator.java +1 −1

diff --git a/org.spearce.jgit/src/org/spearce/jgit/revwalk/StartGenerator.java b/org.spearce.jgit/src/org/spearce/jgit/revwalk/StartGenerator.java
index 7ddcd3c..1b7947f 100644
--- a/org.spearce.jgit/src/org/spearce/jgit/revwalk/StartGenerator.java
+++ b/org.spearce.jgit/src/org/spearce/jgit/revwalk/StartGenerator.java
@@ -144,7 +144,7 @@ RevCommit next() throws MissingObjectException,
 				&& (g.outputType() & SORT_TOPO) == 0)
 			g = new TopoSortGenerator(g);
 		if (walker.hasRevSort(RevSort.REVERSE))
-			g = new LIFORevQueue(q);
+			g = new LIFORevQueue(g);
 		if (boundary)
 			g = new BoundaryGenerator(w, g);

-- 
1.6.0.1.319.g9f32b

Shawn O. Pearce· Sep 7, 2008, 21:03 UTC · re: Shawn O. Pearce · lore

Re: [JGIT PATCH] RevWalk: Fix RevSort.REVERSE combined with RevSort.TOPO

"Shawn O. Pearce" <spearce@spearce.org> wrote:

Show 7 quoted lines

> From: Shawn O. Pearce <sop@google.com>
> 
> The reverse sort was using the initial pending queue, not the
> generator that comes before it in the workflow.  This caused
> a combination of TOPO and REVERSE to return no commits.
> 
> Signed-off-by: Shawn O. Pearce <sop@google.com>

Arrrgh.  Wrong email.

Drop the From and fix the SOB to be:

Signed-off-by: Shawn O. Pearce <spearce@spearce.org>

-- 
Shawn.

Robin Rosenberg· Sep 7, 2008, 21:21 UTC · re: Shawn O. Pearce · lore

Re: [JGIT PATCH] RevWalk: Fix RevSort.REVERSE combined with RevSort.TOPO

söndagen den 7 september 2008 23.00.08 skrev Shawn O. Pearce:

Show 5 quoted lines

> From: Shawn O. Pearce <sop@google.com>
> 
> The reverse sort was using the initial pending queue, not the
> generator that comes before it in the workflow.  This caused
> a combination of TOPO and REVERSE to return no commits.

Some unit tests to back this up would be nice. We running low on coverage in this section of code.

> Signed-off-by: Shawn O. Pearce <sop@google.com>

No idea Google employees had this short email addresses. Ok, 27x27x27 = 19683 would be sufficient in theory, I'd suppose.

-- robin

Shawn O. Pearce· Sep 7, 2008, 21:32 UTC · re: Robin Rosenberg · lore

Re: [JGIT PATCH] RevWalk: Fix RevSort.REVERSE combined with RevSort.TOPO

Robin Rosenberg <robin.rosenberg.lists@dewire.com> wrote:

Show 9 quoted lines

> söndagen den 7 september 2008 23.00.08 skrev Shawn O. Pearce:
> > From: Shawn O. Pearce <sop@google.com>
> > 
> > The reverse sort was using the initial pending queue, not the
> > generator that comes before it in the workflow.  This caused
> > a combination of TOPO and REVERSE to return no commits.
> 
> Some unit tests to back this up would be nice. We running low on
> coverage in this section of code.

Yea, I know. Its way back on my things-to-do list. Right now I'm a bit too swamped with day-job to try and put unit tests into the revwalk package. There's two open issues on the issue tracker that tests should be able to cover and permit debugging of.

I'll come back to revwalk tests. I did some tests for treewalk not too long ago, but even more are still need there too. And lets not even mention the transport package. ;-)

I ran into this bug because I was relying on RevWalk in my day-job
project.  When it barfed I fixed it, and moved on.

> > Signed-off-by: Shawn O. Pearce <sop@google.com>
> 
> No idea Google employees had this short email addresses.  Ok, 27x27x27 = 19683 
> would be sufficient in theory, I'd suppose.

Yea, you can get a short address if you ask nicely. Given that everything is email based its nice having a short address. Makes it easy to tell others in person how to reach me. Fortunately most of the company uses longer addresses, so three (and four) letter words are still available.

I must not have my egit.git repository configured right on my work laptop. I usually don't do egit work there.

-- 
Shawn.