Re: Diffs "from" working directory

Re: Diffs "from" working directory

[J. Bruce Fields]

On Sun, Nov 20, 2005 at 06:17:15PM +0100, Blaisorblade wrote:
> I've given a look at commit fcc1ad70be5ff89996358b6d770cd229fec2a3ba, and 
> while it may or may not be the right thing to do, there's a misunderstanding 
> at the base.

What misunderstanding?  Could you quote it?

> For instance, in the stg tree, before that patch, I can do:
> 
> $ echo ciao >> README
> $ stg diff -r HEAD:
> $ stg diff -r HEAD
> diff --git a/README b/README
> index 184ded8..adb3970 100644
> --- a/README
> +++ b/README
> @@ -12,3 +12,4 @@ StGIT commands.
> 
>  For the latest version see http://www.procode.org/stgit/
>  For a tutorial see http://wiki.procode.org/cgi-bin/wiki.cgi/StGIT_Tutorial
> +ciao

That's a diff *to* the working directory.  My complaint was that there
wasn't a convenient way to get a diff *from* the working directory.

It's nice to have a convenient shorthand for the current working
directory, and since there's already a shorthand for the HEAD ("HEAD")
it seemed more sensible to make "" the current working directory.

It's a minor point, but it seems convenient.

--b.

Re: Diffs "from" working directory

[Linus Torvalds]

On Sun, 20 Nov 2005, J. Bruce Fields wrote:
> 
> That's a diff *to* the working directory.  My complaint was that there
> wasn't a convenient way to get a diff *from* the working directory.

Something like

	git diff -R -M -p

should do it, no?

("git diff" has the strange property that if you give it any diff flags at 
all, it forgets all the default ones. So it defaults to "-M -p", but if 
you just say "-R", it will drop its own defaults, and use _just_ "-R").

[ Which is almost certainly not what most users want, but it's practical 
  when you want to get the internal git format. Whether that is important 
  enough to confuse users or not, I dunno ]

		Linus

Re: Diffs "from" working directory

[J. Bruce Fields]

On Sun, Nov 20, 2005 at 10:13:32AM -0800, Linus Torvalds wrote:
> On Sun, 20 Nov 2005, J. Bruce Fields wrote:
> > 
> > That's a diff *to* the working directory.  My complaint was that there
> > wasn't a convenient way to get a diff *from* the working directory.
> 
> Something like
> 
> 	git diff -R -M -p
>
> should do it, no?

A "-R" option to "stg diff" would be convenient, sure.--b.

Re: Diffs "from" working directory

[Chuck Lever]

[-- Attachment #1: Type: text/plain, Size: 660 bytes --]

J. Bruce Fields wrote:
> On Sun, Nov 20, 2005 at 10:13:32AM -0800, Linus Torvalds wrote:
> 
>>On Sun, 20 Nov 2005, J. Bruce Fields wrote:
>>
>>>That's a diff *to* the working directory.  My complaint was that there
>>>wasn't a convenient way to get a diff *from* the working directory.
>>
>>Something like
>>
>>	git diff -R -M -p
>>
>>should do it, no?
> 
> 
> A "-R" option to "stg diff" would be convenient, sure.--b.

that might be an even more intuitive way to dig out what is wanted.

btw, catalin, this was bruce's patch.  i'm not sure why i was listed as 
the author (probably a mistake of mine when i imported his patch into my 
repository).  ah well.

[-- Attachment #2: cel.vcf --]
[-- Type: text/x-vcard, Size: 439 bytes --]

begin:vcard
fn:Chuck Lever
n:Lever;Charles
org:Network Appliance, Incorporated;Linux NFS Client Development
adr:535 West William Street, Suite 3100;;Center for Information Technology Integration;Ann Arbor;MI;48103-4943;USA
email;internet:cel@citi.umich.edu
title:Member of Technical Staff
tel;work:+1 734 763-4415
tel;fax:+1 734 763 4434
tel;home:+1 734 668-1089
x-mozilla-html:FALSE
url:http://www.monkey.org/~cel/
version:2.1
end:vcard

Re: Diffs "from" working directory

[Catalin Marinas]

On 21/11/05, Chuck Lever <cel@citi.umich.edu> wrote:
> J. Bruce Fields wrote:
> > A "-R" option to "stg diff" would be convenient, sure.--b.
>
> that might be an even more intuitive way to dig out what is wanted.

I also like the idea of having a -R option (--reverse the long
version). This would mean reversing the commit that changed the diff
-r option.

> btw, catalin, this was bruce's patch.  i'm not sure why i was listed as
> the author (probably a mistake of mine when i imported his patch into my
> repository).  ah well.

My import command sets the author to the e-mail sender, which was you.
Maybe this should be changed but I don't know which option is better.
In the meantime, you can change the default e-mail template to set the
From: line with to the author of the patch and maybe add a Reply-to:
with your address.

--
Catalin

Re: Diffs "from" working directory

[Junio C Hamano]

Catalin Marinas <catalin.marinas@gmail.com> writes:

> My import command sets the author to the e-mail sender, which was you.
> Maybe this should be changed but I don't know which option is better.
> In the meantime, you can change the default e-mail template to set the
> From: line with to the author of the patch and maybe add a Reply-to:
> with your address.

... and Sender: perhaps?

The first lines in the commit log message sent over e-mail can
have "Subject: " (or "[PATCH] "), "From: ", and "Date: " to
override what is slurped from e-mail headers if you use mailinfo,
so that might be a better alternative to suggest.

Re: Diffs "from" working directory

[Catalin Marinas]

On 21/11/05, Junio C Hamano <junkio@cox.net> wrote:
> Catalin Marinas <catalin.marinas@gmail.com> writes:
>
> > My import command sets the author to the e-mail sender, which was you.
> > Maybe this should be changed but I don't know which option is better.
> > In the meantime, you can change the default e-mail template to set the
> > From: line with to the author of the patch and maybe add a Reply-to:
> > with your address.
>
> ... and Sender: perhaps?

Yes.

> The first lines in the commit log message sent over e-mail can
> have "Subject: " (or "[PATCH] "), "From: ", and "Date: " to
> override what is slurped from e-mail headers if you use mailinfo,
> so that might be a better alternative to suggest.

Thanks for the suggestion. It is indeed a good idea to parse the
commit log and override the information from the headers.

--
Catalin

Re: Diffs "from" working directory

[Chuck Lever]

[-- Attachment #1: Type: text/plain, Size: 675 bytes --]

Catalin Marinas wrote:
>>btw, catalin, this was bruce's patch.  i'm not sure why i was listed as
>>the author (probably a mistake of mine when i imported his patch into my
>>repository).  ah well.
> 
> 
> My import command sets the author to the e-mail sender, which was you.
> Maybe this should be changed but I don't know which option is better.
> In the meantime, you can change the default e-mail template to set the
> From: line with to the author of the patch and maybe add a Reply-to:
> with your address.

for some reason i was under the impression that it would parse the 
Signed-off-by: fields in the patch description, and take the first one 
as the patch author.

[-- Attachment #2: cel.vcf --]
[-- Type: text/x-vcard, Size: 439 bytes --]

begin:vcard
fn:Chuck Lever
n:Lever;Charles
org:Network Appliance, Incorporated;Linux NFS Client Development
adr:535 West William Street, Suite 3100;;Center for Information Technology Integration;Ann Arbor;MI;48103-4943;USA
email;internet:cel@citi.umich.edu
title:Member of Technical Staff
tel;work:+1 734 763-4415
tel;fax:+1 734 763 4434
tel;home:+1 734 668-1089
x-mozilla-html:FALSE
url:http://www.monkey.org/~cel/
version:2.1
end:vcard

Re: Diffs "from" working directory

[Linus Torvalds]

On Tue, 22 Nov 2005, Chuck Lever wrote:
> 
> for some reason i was under the impression that it would parse the
> Signed-off-by: fields in the patch description, and take the first one as the
> patch author.

The first sign-off really isn't necessarily the author.

It might be a company sign-off (many companies don't want any random 
engineer to send out patches), but much more commonly it's a trivial patch 
that somebody else signs off on, even if the original patcher didn't (see 
case (b) in the sign-off-rules: you can sign of on somebody elses work if 
you know it's under the GPL).

So the fact that there was a sign-off procedure doesn't automatically mean 
that the author will be the first sign-off person, although in _practice_ 
that obviously would likely always be the most common case by far.

(Another reason is that some people actually add the sign-offs above 
previous ones. It happens, although if I notice, I try to point it out).

So authorship really is totally separate from sign-off, and all _my_ tools 
take the authorship from the first "From:" line at the top of the message 
body or from the email itself.

			Linus

Re: Diffs "from" working directory

[Catalin Marinas]

On 22/11/05, Chuck Lever <cel@citi.umich.edu> wrote:
> Catalin Marinas wrote:
> > My import command sets the author to the e-mail sender, which was you.
>
> for some reason i was under the impression that it would parse the
> Signed-off-by: fields in the patch description, and take the first one
> as the patch author.

If you import a patch file and don't specify an author (and the patch
is not an e-mail), the first Signed-off-by: line is used. As Linus
mentioned, we could get rid of this assumption entirely and just
report an error if no author information is given.

--
Catalin

Re: Diffs "from" working directory

[Chuck Lever]

[-- Attachment #1: Type: text/plain, Size: 1670 bytes --]

Linus Torvalds wrote:
> 
> On Tue, 22 Nov 2005, Chuck Lever wrote:
> 
>>for some reason i was under the impression that it would parse the
>>Signed-off-by: fields in the patch description, and take the first one as the
>>patch author.
> 
> 
> The first sign-off really isn't necessarily the author.
> 
> It might be a company sign-off (many companies don't want any random 
> engineer to send out patches), but much more commonly it's a trivial patch 
> that somebody else signs off on, even if the original patcher didn't (see 
> case (b) in the sign-off-rules: you can sign of on somebody elses work if 
> you know it's under the GPL).

heh.  in fact that is what my company (NetApp) requires.

> So the fact that there was a sign-off procedure doesn't automatically mean 
> that the author will be the first sign-off person, although in _practice_ 
> that obviously would likely always be the most common case by far.
> 
> (Another reason is that some people actually add the sign-offs above 
> previous ones. It happens, although if I notice, I try to point it out).
> 
> So authorship really is totally separate from sign-off, and all _my_ tools 
> take the authorship from the first "From:" line at the top of the message 
> body or from the email itself.

then perhaps the problem is that the "stg mail" tool should place the 
author in the From: field automatically?  (ie change the tool, or 
permanently modify the default template that comes with StGIT to do 
this, as Catalin suggested earlier).

that seems a little twisty to me; you're overloading the SMTP header 
field instead of explicitly specifying patch authorship.  seems like a 
layering violation.

[-- Attachment #2: cel.vcf --]
[-- Type: text/x-vcard, Size: 439 bytes --]

begin:vcard
fn:Chuck Lever
n:Lever;Charles
org:Network Appliance, Incorporated;Linux NFS Client Development
adr:535 West William Street, Suite 3100;;Center for Information Technology Integration;Ann Arbor;MI;48103-4943;USA
email;internet:cel@citi.umich.edu
title:Member of Technical Staff
tel;work:+1 734 763-4415
tel;fax:+1 734 763 4434
tel;home:+1 734 668-1089
x-mozilla-html:FALSE
url:http://www.monkey.org/~cel/
version:2.1
end:vcard

Re: Diffs "from" working directory

[Linus Torvalds]

On Tue, 22 Nov 2005, Chuck Lever wrote:
> 
> then perhaps the problem is that the "stg mail" tool should place the author
> in the From: field automatically?  (ie change the tool, or permanently modify
> the default template that comes with StGIT to do this, as Catalin suggested
> earlier).
> 
> that seems a little twisty to me; you're overloading the SMTP header field
> instead of explicitly specifying patch authorship.  seems like a layering
> violation.

No, I only use the actual SMTP header field if the _body_ of the email 
doesn't contain the "From:".

So there's really two different "From:" lines: there's the SMTP header 
one, which is just a default fallback one, and there's the first non-empty 
line of the email body itself, which is the preferred one. No layering 
violation, just two different layers that have the same format for the 
line.

See "The Perfect Patch" by Andrew, and bullet (4): Attribution:

	http://www.zip.com.au/~akpm/linux/patches/stuff/tpp.txt

To quote:

   'If someone else wrote the patch, they should be credited (and blamed) 
    for it. To communicate this, add a line:

    From: John Doe <jdoe@wherever.com>

    as the very first line of the email.  Downstream tools will pick this 
    up and jdoe will get the git "Author" line.'

and I'd be even more anal about it: I would seriously suggest to people 
that they just _always_ add the "From:" line at the head of the email, 
even if it just is exactly the same as what will be in the SMTP header.

Why? Simple. It makes is less likely that somebody who just forwards the 
patch will forget to add that line for you. So you are really helping 
people out - and making sure the attribution stays correct - by adding 
that extra "From:" line at the top of your email body, even if it is 
"unnecessary" in the sense that it's also in your SMTP header.

		Linus