git and (p9p) acme

git and (p9p) acme

[Mathieu Lonjaret]

Hello,

Is anyone using git with p9p acme set as the editor?
Most things here work fine (as long as the pager is set to cat or
something like that), but I've been having problems with 'git commit
--amend'. If I use B as the $EDITOR, git considers the log file edited
as soon as it is opened in acme (and does not wait for a Put), and if
I use E instead (which I thought was meant for that), git does not
"notice" when I Put the file, so it just hangs there.
Am I missing something? is there a simple solution to that?

Cheers,
Mathieu

Re: [9fans] git and (p9p) acme

[dexen deVries]

hi list,


On Thursday 19 of April 2012 13:51:40 Mathieu Lonjaret wrote:
> Is anyone using git with p9p acme set as the editor?

i have the following in ~dexen/.profile
export VISUAL=E
if [ "$DISPLAY" ]; then
        nohup plumber </dev/null >/dev/null 2>/dev/null &
fi


plus a few custom rules in ~dexen/lib/plumbing to deal with PHP's `FILE on
line LINE' convention in error messages etc.


> Most things here work fine (as long as the pager is set to cat or
> something like that), but I've been having problems with 'git commit
> --amend'. If I use B as the $EDITOR, git considers the log file edited
> as soon as it is opened in acme (and does not wait for a Put), and if
> I use E instead (which I thought was meant for that), git does not
> "notice" when I Put the file, so it just hangs there.
> Am I missing something? is there a simple solution to that?

i use `git gui' for commits, but had similar problem with `git rebase -i'. git
did not notice change in file when i merely re-ordered commits, cause the total
length stays the same, and mtime somehow doesn't get listed by `ls -l'.

git notices change of file when the editor exit()s. E has a simple loop around
`ls -l $1' to detect file change; exit()s when it notices change in ls' output.
if you have strange output of `ls', E may not notice change.

a possible workaround is to change length of the file -- add something  in the
comment (hashed-out) section of commit message. in case of `git rebase -i',
change `edit' to `e'.

or re-implement E, really.

caveat:
using VISUAL=E with git, be wary of saving the file several times; the first Put
will cause git to go ahead and subsequent changes to file will most likely be
ignored.


cheers,
--
dexen deVries


Weightless and alone
you speed through the eerie nothingness of space
you circle 'round the Moon
and journey back
to face the punishing torment of re-entry

-- LUNA-C, ``Supaset8 (full release)'', #24m52s

Re: [9fans] git and (p9p) acme

[Aram Hăvărneanu]

dexen deVries wrote:
> caveat:
> using VISUAL=E with git, be wary of saving the file several times; the first Put
> will cause git to go ahead and subsequent changes to file will most likely be
> ignored.

Great caveat, it bit me once.

-- 
Aram Hăvărneanu

Re: [9fans] git and (p9p) acme

[Aram Hăvărneanu]

> Is anyone using git with p9p acme set as the editor?
> Most things here work fine (as long as the pager is set to cat or
> something like that), but I've been having problems with 'git commit
> --amend'. If I use B as the $EDITOR, git considers the log file edited
> as soon as it is opened in acme (and does not wait for a Put), and if
> I use E instead (which I thought was meant for that), git does not
> "notice" when I Put the file, so it just hangs there.
> Am I missing something? is there a simple solution to that?

I do. Rather than remembering what I did, I'll link to my environment:

  http://code.google.com/p/aram-dotfiles/source/browse/

Check the big Plan 9 branch in profile:

  http://code.google.com/p/aram-dotfiles/source/browse/profile

-- 
Aram Hăvărneanu

Re: git and (p9p) acme

[sqweek]

On 19 April 2012 19:51, Mathieu Lonjaret <mathieu.lonjaret@gmail.com> wrote:
> Is anyone using git with p9p acme set as the editor?
<snip>
> I use E instead (which I thought was meant for that), git does not
> "notice" when I Put the file, so it just hangs there.

 I never tried using EDITOR=E, instead wrote my own little script. I'm
a little compulsive when it comes to hitting Put, so my script
actually waits until you delete the acme window. Also you can hit
"Cancel" to have the script exit non-zero. Anyway this is what I use
for $EDITOR/$VISUAL:

http://sqweek.net/plan9/acmeedit

 Also, sup! It's been awhile. You still rock climbing? :)
-sqweek

Re: git and (p9p) acme

[Mathieu Lonjaret]

Thanks Dexen and Aram for the explanation.

On Fri, Apr 20, 2012 at 6:04 PM, sqweek <sqweek@gmail.com> wrote:
> On 19 April 2012 19:51, Mathieu Lonjaret <mathieu.lonjaret@gmail.com> wrote:
>> Is anyone using git with p9p acme set as the editor?
> <snip>
>> I use E instead (which I thought was meant for that), git does not
>> "notice" when I Put the file, so it just hangs there.
>
>  I never tried using EDITOR=E, instead wrote my own little script. I'm
> a little compulsive when it comes to hitting Put, so my script
> actually waits until you delete the acme window. Also you can hit
> "Cancel" to have the script exit non-zero. Anyway this is what I use
> for $EDITOR/$VISUAL:
>
> http://sqweek.net/plan9/acmeedit

Nice, it almost works out of the box.
I got a 9p error message when using it:
9p: write error: ill-formed control message
but it actually does the job anyway.
Thanks!

I'll investigate later on to see which part fails exactly.

Re: git and (p9p) acme

[Mark van Atten]

Mathieu Lonjaret <mathieu.lonjaret@gmail.com> a écrit :

>> http://sqweek.net/plan9/acmeedit
>
> Nice, it almost works out of the box.
> I got a 9p error message when using it:
> 9p: write error: ill-formed control message

In /lib/acme.rc, there is

fn newwindow {
         winctl=`{9p read acme/new/ctl}
         winid=$winctl(1)
         winctl noscroll
}

I think the noscroll no longer functions; I commented it out, and then  
acmeedit no longer produces the error message.

Best wishes,
Mark.

-- 
Directeur de recherches, CNRS
IHPST, 13 rue du Four, 75006 Paris, France
tel ++ 33 (0)1 43 54 60 36, fax ++ 33 (0)1 43 25 29 48
http://www-ihpst.univ-paris1.fr/en/6,mark_van_atten.html


-- 
Ce message a ete verifie par MailScanner
pour des virus ou des polluriels et rien de
suspect n'a ete trouve.

Re: git and (p9p) acme

[Mathieu Lonjaret]

Indeed.

On Fri, Apr 20, 2012 at 9:39 PM, Mark van Atten
<Mark.vanAtten@univ-paris1.fr> wrote:
> Mathieu Lonjaret <mathieu.lonjaret@gmail.com> a écrit :
>
>
>>> http://sqweek.net/plan9/acmeedit
>>
>>
>> Nice, it almost works out of the box.
>> I got a 9p error message when using it:
>> 9p: write error: ill-formed control message
>
>
> In /lib/acme.rc, there is
>
> fn newwindow {
>        winctl=`{9p read acme/new/ctl}
>        winid=$winctl(1)
>        winctl noscroll
> }
>
> I think the noscroll no longer functions; I commented it out, and then
> acmeedit no longer produces the error message.
>
> Best wishes,
> Mark.
>
> --
> Directeur de recherches, CNRS
> IHPST, 13 rue du Four, 75006 Paris, France
> tel ++ 33 (0)1 43 54 60 36, fax ++ 33 (0)1 43 25 29 48
> http://www-ihpst.univ-paris1.fr/en/6,mark_van_atten.html
>
>
> --
> Ce message a ete verifie par MailScanner
> pour des virus ou des polluriels et rien de
> suspect n'a ete trouve.
>

Re: git and (p9p) acme

[Russ Cox]

I guess it depends on what you expect to break.
I like using the E script because (once you learn not
to click Put, which doesn't take too long) after I click
Put the window is still there with my work in it in case
the tool I just sent it to chooses to blow up and
discard my efforts.

Russ

Re: [9fans] git and (p9p) acme

[Lyndon Nerenberg]

On 2012-04-25, at 1:21 PM, Russ Cox wrote:

> I like using the E script because (once you learn not
> to click Put, which doesn't take too long) after I click
> Put the window is still there with my work in it in case
> the tool I just sent it to chooses to blow up and
> discard my efforts.

If 'sam file' did the obvious thing this would be much less of a problem.

Re: [9fans] git and (p9p) acme

[Russ Cox]

On Wed, Apr 25, 2012 at 4:30 PM, Lyndon Nerenberg <lyndon@orthanc.ca> wrote:
> If 'sam file' did the obvious thing this would be much less of a problem.

What is not obvious about what 'sam file' does?

Re: [9fans] git and (p9p) acme

[Lyndon Nerenberg]

On 2012-04-25, at 1:36 PM, Russ Cox wrote:

> What is not obvious about what 'sam file' does?

Plugging 'sam file' into a script does not launch the editor with the specified file in a window for the user to edit, and then save out.

People are clamouring for a (visual) 'ed foo' replacement.  Acme can't do that.  Sam can, but it needs some unintuitive arguments.  If 'E foo' wrapped sam up in the proper glue to make it fake out 'vi foo' as far as the atomic editing from the command line needs that started this thread, that would be wonderful.

That aside, I think rio+acme provides enough glue to make it possible to have an 'E' script that launches an acme window and blocks until the window is closed.  If I have time over the weekend I'll fiddle about and see what comes out.

Re: [9fans] git and (p9p) acme

[John Floren]

On Wed, Apr 25, 2012 at 1:46 PM, Lyndon Nerenberg <lyndon@orthanc.ca> wrote:
>
> On 2012-04-25, at 1:36 PM, Russ Cox wrote:
>
>> What is not obvious about what 'sam file' does?
>
> Plugging 'sam file' into a script does not launch the editor with the specified file in a window for the user to edit, and then save out.

Works for me... this is what I do when writing commit messages in hg.
Sure, you have to go into the right-button menu and open the file
window, but it's 2 mouse clicks and I find sam a lot more convenient
than acme for this kind of quick edit.

What behavior are you seeing?

john

Re: [9fans] git and (p9p) acme

[Lyndon Nerenberg]

On 2012-04-25, at 1:51 PM, John Floren wrote:

> What behavior are you seeing?

A lack of expectation.  I.e. I expect 'sam foo' to, well, edit 'foo'.  'acme foo' does that, as does 'ed foo.'  From a UI perspective, the 'sam foo' behaviour is as non-intuitive as it gets.

If 'sam foo' (i.e. with a single filename argument) loaded the file into the edit window, it would mimic the behaviour of almost every other editor out there.

--lyndon

Re: [9fans] git and (p9p) acme

[Russ Cox]

On Wed, Apr 25, 2012 at 4:57 PM, Lyndon Nerenberg <lyndon@orthanc.ca> wrote:
> A lack of expectation.  I.e. I expect 'sam foo' to, well, edit 'foo'.  'acme foo' does that, as does 'ed foo.'  From a UI perspective, the 'sam foo' behaviour is as non-intuitive as it gets.
>
> If 'sam foo' (i.e. with a single filename argument) loaded the file into the edit window, it would mimic the behaviour of almost every other editor out there.

Are you complaining that 'sam foo' pops up a sam window and
you have to click to place the new window for foo instead of having
the window automatically placed?

Otherwise, I don't know what you're talking about.
sam foo starts a copy of sam editing foo.

Russ

Re: [9fans] git and (p9p) acme

[Lyndon Nerenberg]

On 2012-04-25, at 1:58 PM, Russ Cox wrote:

> Otherwise, I don't know what you're talking about.

Yes you do.

Re: [9fans] git and (p9p) acme

[Russ Cox]

On Wed, Apr 25, 2012 at 4:46 PM, Lyndon Nerenberg <lyndon@orthanc.ca> wrote:
> Plugging 'sam file' into a script does not launch the editor with the specified file in a window for the user to edit, and then save
out.

Sure it does.  And so does 'acme file', if you are on Plan 9 or
if you are on plan9port and don't already have acme running.

The point of the E script is to let you edit in the editor you're
already using instead of opening a new one.

Russ

Re: [9fans] git and (p9p) acme

[Lyndon Nerenberg]

On 2012-04-25, at 1:56 PM, Russ Cox wrote:

> The point of the E script is to let you edit in the editor you're
> already using instead of opening a new one.

But none of those commands will block until the file is saved, and that's the issue.

Re: [9fans] git and (p9p) acme

[Russ Cox]

On Wed, Apr 25, 2012 at 4:59 PM, Lyndon Nerenberg <lyndon@orthanc.ca> wrote:
> But none of those commands will block until the file is saved, and that's the issue.

Are you sure you are running 'sam foo' and not 'sam foo &'?
Sam 'blocks' until you quit out of it, just like acme or vi or ed.

Russ

Re: [9fans] git and (p9p) acme

[Lyndon Nerenberg]

On 2012-04-25, at 2:02 PM, Russ Cox wrote:

> Are you sure you are running 'sam foo' and not 'sam foo &'?
> Sam 'blocks' until you quit out of it, just like acme or vi or ed.

It doesn't load the file for the user to edit.  Given a single file on the command line, the obvious behaviour should be to load the file the user asked to edit.

Re: [9fans] git and (p9p) acme

[Lyndon Nerenberg]

On 2012-04-25, at 2:02 PM, Russ Cox wrote:

> Are you sure you are running 'sam foo' and not 'sam foo &'?
> Sam 'blocks' until you quit out of it, just like acme or vi or ed.

And you're right -- blocking behaviour isn't the problem with sam.

Re: [9fans] git and (p9p) acme

[Russ Cox]

> It doesn't load the file for the user to edit.

You must be running a different sam than I am.
I would make a video and post it on YouTube
but I am too lazy.

Russ

Re: git and (p9p) acme

[Mathieu Lonjaret]

So, are you saying E just works out of the box for you as the $EDITOR
for git/hg?
Or did you have to tweak a few things like what Aram and Dexen describe?

I wouldn't mind using E as it is, with Put being the trigger to the
tool (git in that case), but it does not seem to be working here.

On Wed, Apr 25, 2012 at 10:21 PM, Russ Cox <rsc@swtch.com> wrote:
> I guess it depends on what you expect to break.
> I like using the E script because (once you learn not
> to click Put, which doesn't take too long) after I click
> Put the window is still there with my work in it in case
> the tool I just sent it to chooses to blow up and
> discard my efforts.
>
> Russ

Re: [9fans] git and (p9p) acme

[Aram Hăvărneanu]

> So, are you saying E just works out of the box for you as the $EDITOR
> for git/hg?
> Or did you have to tweak a few things like what Aram and Dexen describe?

I guess all you need to run up awd in your cd alias. I think I stole
my cd function from Russ.

-- 
Aram Hăvărneanu

Re: [9fans] git and (p9p) acme

[Aram Hăvărneanu]

> I guess all you need to run up awd in your cd alias. I think I stole
> my cd function from Russ.

Actually, the way one does this is documented in label(1). Beware that
this requires X, hence the $DISPLAY check in my profile.

-- 
Aram Hăvărneanu

Re: [9fans] git and (p9p) acme

[Mathieu Lonjaret]

Indeed, it seems like awd was all that was needed.
Thanks!

On Wed, Apr 25, 2012 at 11:49 PM, Aram Hăvărneanu <aram.h@mgk.ro> wrote:
>> I guess all you need to run up awd in your cd alias. I think I stole
>> my cd function from Russ.
>
> Actually, the way one does this is documented in label(1). Beware that
> this requires X, hence the $DISPLAY check in my profile.
>
> --
> Aram Hăvărneanu
>

Re: git and (p9p) acme

[Russ Cox]

On Wed, Apr 25, 2012 at 5:35 PM, Mathieu Lonjaret
<mathieu.lonjaret@gmail.com> wrote:
> So, are you saying E just works out of the box for you as the $EDITOR
> for git/hg?

Yes.  Note that E waits for the file to change by running
ls -l on the file in a loop.  If for some reason doing the Put
does not change the ls -l output (date+size, primarily) then
E will not exit.  If you are having trouble I would suggest
putting some echo statements in the script or just doing
'set -e' at the top so that it echoes each statement as it
executes (including the all-important comparison).

I cannot imagine why using awd would change the behavior
of the E script.

Russ

Re: [9fans] git and (p9p) acme

[Rob Pike]

set -x

-rob

Re: git and (p9p) acme

[Mathieu Lonjaret]

I've just retried on the machine where I had the problem and it now
just works. No need even for awd indeed.
I'll try and reproduce the behavior, but I just don't see what I may
have done different the previous times...

thanks,
Mathieu

On Thu, Apr 26, 2012 at 5:32 AM, Russ Cox <rsc@swtch.com> wrote:
> On Wed, Apr 25, 2012 at 5:35 PM, Mathieu Lonjaret
> <mathieu.lonjaret@gmail.com> wrote:
>> So, are you saying E just works out of the box for you as the $EDITOR
>> for git/hg?
>
> Yes.  Note that E waits for the file to change by running
> ls -l on the file in a loop.  If for some reason doing the Put
> does not change the ls -l output (date+size, primarily) then
> E will not exit.  If you are having trouble I would suggest
> putting some echo statements in the script or just doing
> 'set -e' at the top so that it echoes each statement as it
> executes (including the all-important comparison).
>
> I cannot imagine why using awd would change the behavior
> of the E script.
>
> Russ

Re: git and (p9p) acme

[Ethan Burns]

[-- Attachment #1.1: Type: text/plain, Size: 1729 bytes --]

Hi everyone,

I wasn't aware of B and E before reading this thread.  I tried using E on
my install but it seems that I don't have a plumb file.  When I use it on a
file, I don't get an acme window and instead I get:

$ touch test
$ E test
plumb: can't open plumb file: dial unix!/tmp/ns.eaburns.:0/plumb: connect
/tmp/ns.eaburns.:0/plumb: No such file or directory
editing test

In /tmp/ns.eaburns.:0/ the only file is called acme.  Do I need to do
something first to create the plumb file or do I need to do something to
tell E to use the acme file instead?

Best,
Ethan


On Thursday, April 26, 2012 6:20:49 PM UTC-4, mpl wrote:
>
> I've just retried on the machine where I had the problem and it now
> just works. No need even for awd indeed.
> I'll try and reproduce the behavior, but I just don't see what I may
> have done different the previous times...
>
> thanks,
> Mathieu
>
> On Thu, Apr 26, 2012 at 5:32 AM, Russ Cox <rsc@swtch.com> wrote:
> > On Wed, Apr 25, 2012 at 5:35 PM, Mathieu Lonjaret
> > <mathieu.lonjaret@gmail.com> wrote:
> >> So, are you saying E just works out of the box for you as the $EDITOR
> >> for git/hg?
> >
> > Yes.  Note that E waits for the file to change by running
> > ls -l on the file in a loop.  If for some reason doing the Put
> > does not change the ls -l output (date+size, primarily) then
> > E will not exit.  If you are having trouble I would suggest
> > putting some echo statements in the script or just doing
> > 'set -e' at the top so that it echoes each statement as it
> > executes (including the all-important comparison).
> >
> > I cannot imagine why using awd would change the behavior
> > of the E script.
> >
> > Russ
>

[-- Attachment #1.2: Type: text/html, Size: 2261 bytes --]

Re: git and (p9p) acme

[Ethan Burns]

[-- Attachment #1.1: Type: text/plain, Size: 577 bytes --]

Hi everyone,

Sorry if this is a double post but I tried to send this before and it said
that my post was deleted for some reason.

I was unaware of E until reading this thread.  When I try using E with my
p9p install I don't get a new acme window, instead I just get this error:

$ E test
plumb: can't open plumb file: dial unix!/tmp/ns.eaburns.:0/plumb: connect
/tmp/ns.eaburns.:0/plumb: No such file or directory
editing test


The only file in /tmp/ns.eaburns.:0/ is called 'acme'.  Is there something
that I need to do to get a plumb file?

Best,
Ethan

[-- Attachment #1.2: Type: text/html, Size: 742 bytes --]

Re: git and (p9p) acme

[Mathieu Lonjaret]

It happened again. and I've just realized it's simply what you had
warned about: since it was an --amend and I didn't change anything in
the message there was no change to the ls -l output, hence E did
nothing when I Put.

On Fri, Apr 27, 2012 at 12:20 AM, Mathieu Lonjaret
<mathieu.lonjaret@gmail.com> wrote:
> I've just retried on the machine where I had the problem and it now
> just works. No need even for awd indeed.
> I'll try and reproduce the behavior, but I just don't see what I may
> have done different the previous times...
>
> thanks,
> Mathieu
>
> On Thu, Apr 26, 2012 at 5:32 AM, Russ Cox <rsc@swtch.com> wrote:
>> On Wed, Apr 25, 2012 at 5:35 PM, Mathieu Lonjaret
>> <mathieu.lonjaret@gmail.com> wrote:
>>> So, are you saying E just works out of the box for you as the $EDITOR
>>> for git/hg?
>>
>> Yes.  Note that E waits for the file to change by running
>> ls -l on the file in a loop.  If for some reason doing the Put
>> does not change the ls -l output (date+size, primarily) then
>> E will not exit.  If you are having trouble I would suggest
>> putting some echo statements in the script or just doing
>> 'set -e' at the top so that it echoes each statement as it
>> executes (including the all-important comparison).
>>
>> I cannot imagine why using awd would change the behavior
>> of the E script.
>>
>> Russ

Re: [9fans] git and (p9p) acme

[dexen deVries]

[-- Attachment #1: Type: text/plain, Size: 636 bytes --]

On Wednesday 25 of April 2012 23:35:59 Mathieu Lonjaret wrote:
> So, are you saying E just works out of the box for you as the $EDITOR
> for git/hg?
> Or did you have to tweak a few things like what Aram and Dexen describe?

worked out-of-box; I only add plumbing for PHP stuff -- and start plumber in
~/.profile

use $VISUAL rather than $EDITOR to ensure it is used only under X Windows

--
dexen deVries



Weightless and alone
you speed through the eerie nothingness of space
you circle 'round the Moon
and journey back
to face the punishing torment of re-entry

-- LUNA-C, ``Supaset8 (full release)'', #24m52s

[-- Attachment #2: plumbing --]
[-- Type: text/plain, Size: 197 bytes --]

#file / line in PHP format
type is text
data matches '(.+) on line ([0-9]+)'
arg isfile $1
data set $file
attr add addr=$2
plumb to edit

include /usr/local/plan9/plumb/initial.plumbing


[-- Attachment #3: dot-profile-fragment --]
[-- Type: text/plain, Size: 100 bytes --]

export VISUAL=E

if [ "$DISPLAY" ]; then
	nohup plumber </dev/null >/dev/null 2>/dev/null &
fi

Re: [9fans] git and (p9p) acme

[Christian Neukirchen]

lyndon@orthanc.ca (Lyndon Nerenberg) writes:

> On 2012-04-25, at 2:02 PM, Russ Cox wrote:
>
>> Are you sure you are running 'sam foo' and not 'sam foo &'?
>> Sam 'blocks' until you quit out of it, just like acme or vi or ed.
>
> It doesn't load the file for the user to edit.

It does, but it doesnt show the buffer by default.  Just right-click and
select it.

--
Christian Neukirchen  <chneukirchen@gmail.com>  http://chneukirchen.org