On Tue, Mar 12, 2013 at 7:19 PM, Terry Reedy email@example.com wrote:
On 3/12/2013 2:52 AM, Ezio Melotti wrote:
What are the exact commands you used?
Clicks on TortoiseHg HgWorkbench GUI ;-).
I wonder if TortoiseHg is doing something wrong here. Maybe you could try from cmd too.
Are your clones up to date (i.e. did you do "hg pull" and "hg up" before "hg graft")?
There were no other pushes between my last de-double patch and this, and I am sure I ran my pull + 3*update .bat first. I have run it multiple times since.
Around the time you pushed on 2.7 I also pushed something, so that might have created some conflict. How does your .bat look like? One gotcha of the share extension is that if you use "hg pull -u" and there's nothing to pull because you already pulled in one of the shared clones, the update won't be executed (this is actually normal behaviour of "hg pull", but the consequences are especially noticeable while using shared clones).
Does "hg heads ." show you more than one head?
The DAG window shows the normal one head per branch as appropriate for the particular branch display. At the moment, hg heads shows the four commits from Eli, 82628 to 82631 as heads plus old 2.6 and 3.1 heads.
Is your clone "clean" (i.e. does "hg status" show anything as 'M')?
The status window is empty until I edit NEWS and click Refresh, at which point M Misc/News shows up with the megadiff. Right click/ Revert/yes and the file is reverted.
Once your clone is clean you can just edit Misc/NEWS manually
Since the graft and import failed (producing no diff), I have been editing manually and that is when I get the megadiff. I added a couple of blank lines to ACKS and got a normal diff. Now, adding a blank line to 2.7 NEWS also gives a blank line.
Could the failed graft have messed up the master copy in my cpython repository.
That's possible. From "hg help graft": If a graft merge results in conflicts, the graft process is interrupted so that the current merge can be manually resolved. Once all conflicts are addressed, the graft process can be continued with the -c/--continue option.
This doesn't mean that you copy is messed up though. "hg up -C 3.2" should restore it. When I graft/merge and there are conflicts I use kdiff3, and it takes just a few seconds to solve the conflicts usually (for Misc/NEWS is ctrl+2, ctrl+3, ctrl+s, alt+f4, that roughly translates too "include both the conflicting news, save and quit).
I have tried deleting the NEWS file and reverting the deletion. hg update does not restore the file as it apparently thinks I actually want the uncommitted deletion.
How did you delete it? I assume that if you do it from the TortoiseHG GUI, it will mark it as "deleted" ('D' in "hg status"). If you do it from cmd/file manager hg should see it as missing ('!' in "hg status") and you can use "hg revert Misc/NEWS" to restore it.
it's easier than trying to graft the 2 changesets you made on 2.7 to add and edit the Misc/NEWS entry.
There was only one 2.7 changeset with only the NEWS patch.
I was referring to the one that added the news + the one that fixed the issue id.
You can also check with "hg in" and "hg out" if there's something you haven't pulled/pushed yet, but that shouldn't be a problem.
I tried both and got 'no changes'.
(If you prefer you can come on #python-dev too.)
I may try that, but I suspect that my registration/nick has expired again and last time is was obnoxiously hard to get re-established.
There's no need to register your nick for #python-dev (there is for #python though). You can just fire up your favourite IRC client (or even http://webchat.freenode.net/) and join. (Registering the nick shouldn't be difficult though.)