how to import tags?

[Augie Fackler]

> On May 24, 2018, at 4:35 PM, Hollis Blanchard <hollis_blanchard at mentor.com> wrote:
> 
> On 05/24/2018 11:38 AM, Augie Fackler wrote:
>> 
>> 
>>> On May 24, 2018, at 2:20 PM, Hollis Blanchard <hollis_blanchard at mentor.com> wrote:
>>> 
>>> On 05/23/2018 03:15 PM, Hollis Blanchard wrote:
>>>> 
>>>> 2. Commit tags as children of the correct head. 
>>>> 
>>>> I don't see a 'tag' cmdline option for this, though: the parent commit is the parent of the working directory, without an ability to override. That means I would need to keep working directories for all these mirrors, and since there are GBs of working directories and millions of files, that would be significant overhead. 
>>>> 
>>> Looks like I can use 'hg debugsetparents' for exactly this purpose:
>>> 
>>> hg debugsetparents REV1 [REV2]
>>> manually set the parents of the current working directory
>>> 
>>>     This is useful for writing repository conversion tools, but should be used
>>>     with care. For example, neither the working directory nor the dirstate is
>>>     updated, so file status may be incorrect after running this command.
>>> I hope that debug command sticks around…
>>> 
>> I’d be very surprised if it went away. I’d probably do what you’re doing given the same constraints.
>> 
> I've run into two complications:
> 
> 1. Bookmarks. If I apply a tag to a bookmarked changeset, I want the bookmark to advance (don't I?). However, debugsetparents doesn't activate the bookmark, so it doesn't advance. To work around that, I could script all the bookmark manipulation by hand, but that's awkward and becoming many python invocations...
> 
> 2. Importing one label become a series of stateful commands: debugsetparents; tag; adjust bookmarks; maybe debugsetparents back. If something interrupts that sequence (e.g. concurrent imports, or even something external like 'kill'), then badness would result.
> 
> It seems like a 'tag --parent' option would solve all of these problems. Is that reasonable, if someone submitted a decent patch for it?
> 

It might, or you could use the cmdserver to avoid the extra Python invocations?

I’m not sure what the semantics of --parent should be.

> Hollis Blanchard
> Mentor Graphics Emulation Division
>