threads / discuss / 12957

"stg reset --status" doesn't just reset status

Subject: "stg reset --status" doesn't just reset status

## tl;dr

3 messages between Apr 2, 2008 and Apr 2, 2008.

replies: 2people: 2as markdown or json

Pavel Roskin· Apr 2, 2008, 02:21 UTC · lore

Hello!

I used to see messages like this:

$ stg pull Checking for changes in the working directory ... done stg pull: local changes in the tree. Use "refresh" or "status --reset"

This time I decided to see what "stg reset --status" actually does, and I was unpleasantly surprised that it would do much more that its name implies.

It doesn't just reset the status (no idea what it would be, but it doesn't sound scary). It removes all local changes. It's essentially "git reset --hard". I can easily imagine that some beginner would lose valuable changes by following that advice while trying to update from the upstream repository.

I would hate to suggest another stg command, as there are too many of them already. On the other hand, if "applied" and "unapplied" are downgraded to switches for "stg series", we probably could justify adding one more command, "stg reset". By the way, the default could be to save the changes to a hidden "stash" patch, and the "--hard" switch would do a real reset.

Another (not alternative) approach would be to have an option to "stg pull" to save the changes as a temporary patch that would be applied and deleted if it applied cleanly. That shouldn't be a default for "stg pull", as it's likely that the user just forgot "stg refresh".

-- 
Regards,
Pavel Roskin

Karl Hasselström· Apr 2, 2008, 12:00 UTC · re: Pavel Roskin · lore

Re: "stg reset --status" doesn't just reset status

On 2008-04-01 22:21:24 -0400, Pavel Roskin wrote:

Show 16 quoted lines

> This time I decided to see what "stg reset --status" actually does,
> and I was unpleasantly surprised that it would do much more that its
> name implies.
>
> It doesn't just reset the status (no idea what it would be, but it
> doesn't sound scary). It removes all local changes. It's essentially
> "git reset --hard". I can easily imagine that some beginner would
> lose valuable changes by following that advice while trying to
> update from the upstream repository.
>
> I would hate to suggest another stg command, as there are too many
> of them already. On the other hand, if "applied" and "unapplied" are
> downgraded to switches for "stg series", we probably could justify
> adding one more command, "stg reset". By the way, the default could
> be to save the changes to a hidden "stash" patch, and the "--hard"
> switch would do a real reset.

(I assume "stg reset --status" is just a typo for "stg status --reset"?)

I'd be fine with removing status --reset, but since there is currently no other way to do this in StGit, I expect Catalin would object. (As I recall, that's precisely what happened when I did try to remove it some time ago.)

I'm currently (slowly) working on an "stg reset" command that'll be able to reset the stack to any prior state. It could be made to reset to the most recent recorded state if no extra argument is given, which I think would make it do what you want.

-- 
Karl Hasselström, kha@treskal.com
      www.treskal.com/kalle

Pavel Roskin· Apr 2, 2008, 17:30 UTC · re: Karl Hasselström · lore

Re: "stg reset --status" doesn't just reset status

On Wed, 2008-04-02 at 14:00 +0200, Karl Hasselström wrote:

> (I assume "stg reset --status" is just a typo for "stg status
> --reset"?)

Exactly. "stg status --reset" is so against the logic that I could not even write it properly :)

> I'd be fine with removing status --reset, but since there is currently
> no other way to do this in StGit, I expect Catalin would object. (As I
> recall, that's precisely what happened when I did try to remove it
> some time ago.)

1) stg diff | patch -Rp1
2) stg new -m "trash" trash; stg refresh; stg delete trash

Besides, I don't think everything should be easily doable from stgit.

> I'm currently (slowly) working on an "stg reset" command that'll be
> able to reset the stack to any prior state. It could be made to reset
> to the most recent recorded state if no extra argument is given, which
> I think would make it do what you want.

Sounds good.  Thank you!

-- 
Regards,
Pavel Roskin