You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@subversion.apache.org by Karl Fogel kf...@newton.ch.collab.net on 2003/01/24 16:19:43 UTC

Issue #1004 and Garrett Rooney

Heh, bet that got your attention, Garrett :-).

We have a lot of issues, but also a lot of developers; I'm trying to
go through all the near issues and match them up with developers.
This is strictly voluntary, of course -- if you can't do it, just say
the word, and we'll move on to the next victim.

Anyway: would you be willing to take on issue #1004?  The summary says

   need api to interrupt long running operations via a client

...but what it's really about is creating a new svn_client_context_t
structure and passing it to all the client functions.  Greg Stein has
more comments in the issue.  Basically, this is something we've needed
for a long time; it will not only give us a way to add cancellation
and progress-meter functionality, but may also simplify the passing of
auth data, the notification functions, etc.

I remember that it's been discussed on this list, and we were all
pretty much for it, while trembling at the thought of actually
implementing it :-).  But you are brave, I know.  Well do I remember
your valour at the Caves of Arnorgaroth, and the flash of your ax as
it bit into your foe's duodenum...

Will you fight by our side again, I say?

-K

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org

Re: [PATCH] svn_client_ctx_t (was Re: Issue #1004 and Garrett Rooney)

Posted by Garrett Rooney ro...@electricjellyfish.net.
Ben Collins-Sussman wrote:

Garrett Rooney rooneg@electricjellyfish.net writes:

  

but that said, your change will likely conflict with large parts of
mine, so if you can wait for me to commit my change (i'll do it
tonight, once i get home from work), that would be nice.  if you're
planning on doing the work today, and will be done by tonight, then
feel free to do it and i'll just merge my changes in, but if you have
other stuff you could do in the meantime, it would simplify my life ;-)
    


Garrett, it turns out I need to do some other non-svn stuff today, so
I won't be able to do my libsvn_auth changes till tomorrow.  So go
ahead and commit your stuff tonight.  :-)


and here i thought i'd never be happy to see ben getting pulled off of 
svn work to do something else ;-)

-garrett


---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org

Re: [PATCH] svn_client_ctx_t (was Re: Issue #1004 and Garrett Rooney)

Posted by Ben Collins-Sussman su...@collab.net.
Garrett Rooney rooneg@electricjellyfish.net writes:

 but that said, your change will likely conflict with large parts of
 mine, so if you can wait for me to commit my change (i'll do it
 tonight, once i get home from work), that would be nice.  if you're
 planning on doing the work today, and will be done by tonight, then
 feel free to do it and i'll just merge my changes in, but if you have
 other stuff you could do in the meantime, it would simplify my life ;-)

Garrett, it turns out I need to do some other non-svn stuff today, so
I won't be able to do my libsvn_auth changes till tomorrow.  So go
ahead and commit your stuff tonight.  :-)

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org

Re: [PATCH] svn_client_ctx_t (was Re: Issue #1004 and Garrett Rooney)

Posted by Garrett Rooney ro...@electricjellyfish.net.
Ben Collins-Sussman wrote:

Garrett Rooney rooneg@electricjellyfish.net writes:

  

On Saturday, January 25, 2003, at 12:16 PM, Karl Fogel wrote:

    

Garrett Rooney rooneg@electricjellyfish.net writes:
      

cause it's easier to just move around existing functionality while i'm
thinking about how to best do the cancelation stuff, and this stuff
should get moved anyway ;-)
        

Gotcha. :-)
      


Hey Garrett... the svn_client_auth_baton_t is about to vanish, and a
plain old svn_auth_baton_t will take its place.  In other words, I'd
like to start making all our libraries use libsvn_auth.  Is your patch
going to conflict with my plan?  Can I get my change in first?  :-)
  


well, i imagine the svn_auth_baton_t will be used similarly to the 
svn_client_auth_baton_t, so it's not a huge deal for me to change it...

but that said, your change will likely conflict with large parts of 
mine, so if you can wait for me to commit my change (i'll do it tonight, 
once i get home from work), that would be nice.  if you're planning on 
doing the work today, and will be done by tonight, then feel free to do 
it and i'll just merge my changes in, but if you have other stuff you 
could do in the meantime, it would simplify my life ;-)

-garrett


---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org

Re: [PATCH] svn_client_ctx_t (was Re: Issue #1004 and Garrett Rooney)

Posted by Ben Collins-Sussman su...@collab.net.
Garrett Rooney rooneg@electricjellyfish.net writes:

 On Saturday, January 25, 2003, at 12:16 PM, Karl Fogel wrote:
 
  Garrett Rooney rooneg@electricjellyfish.net writes:
  cause it's easier to just move around existing functionality while i'm
  thinking about how to best do the cancelation stuff, and this stuff
  should get moved anyway ;-)
 
  Gotcha. :-)

Hey Garrett... the svn_client_auth_baton_t is about to vanish, and a
plain old svn_auth_baton_t will take its place.  In other words, I'd
like to start making all our libraries use libsvn_auth.  Is your patch
going to conflict with my plan?  Can I get my change in first?  :-)

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org

Re: [PATCH] svn_client_ctx_t (was Re: Issue #1004 and Garrett Rooney)

Posted by Garrett Rooney ro...@electricjellyfish.net.
On Saturday, January 25, 2003, at 12:16 PM, Karl Fogel wrote:

 Garrett Rooney rooneg@electricjellyfish.net writes:
 cause it's easier to just move around existing functionality while i'm
 thinking about how to best do the cancelation stuff, and this stuff
 should get moved anyway ;-)

 Gotcha. :-)

another reason to move the existing callbacks and batons into a context 
parameter (if we're going to have a context parameter, which it seems 
like everyone agrees is a good idea) is that getting it over with ASAP 
means we get the API change out of the way now, minimizing the number 
of third party projects that have to deal with it...  at the moment we 
have a small number of consumers outside of the svn command line 
client, but that number is growing quickly.

anyway, here's my current patch.  i haven't run 'make check' lately 
(since changing the property functions), but it most likely works.  i 
don't have the energy to update the log message at the moment, so just 
take the previous one, and extrapolate out to include all the other 
functions in svn_client.h that had auth_baton parameters.

the only other thing i want to do with this (other than testing it 
more) is to possibly add an svn_context_t parameter to 
svn_opt_subcommand_t, since all the svn_client_* functions take one, 
and it seems silly to have the code to create it and set the auth_baton 
in each subcommand.

as always, i'd welcome any input people have, as i'd like to get it 
into the tree soon so that i can move on to work on the other parts of 
this issue.

-garrett


Re: [PATCH] svn_client_ctx_t (was Re: Issue #1004 and Garrett Rooney)

Posted by Karl Fogel kf...@newton.ch.collab.net.
Garrett Rooney rooneg@electricjellyfish.net writes:
 cause it's easier to just move around existing functionality while i'm
 thinking about how to best do the cancelation stuff, and this stuff
 should get moved anyway ;-)

Gotcha. :-)

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org

Re: [PATCH] svn_client_ctx_t (was Re: Issue #1004 and Garrett Rooney)

Posted by Garrett Rooney ro...@electricjellyfish.net.
On Saturday, January 25, 2003, at 12:15 AM, Karl Fogel wrote:

 I don't think the client_ctx_t baton itself would ever get passed down
 into libsvn_wc; but some of the things *in* the baton might be passed
 as individual arguments to libsvn_wc functions, of course.

in retrospect, that does make more sense.

 the first thing i thought would be nice to move was the
 svn_client_auth_baton_t's that get passed to everything that accesses
 the repository directly.  i've got a first cut at that done, just
 doing so for svn_client_checkout (and the functions were affected by
 that change, progressing only far enough to make things build and pass
 make check).

 assuming people like this approach, i was going to finish up moving
 the auth batons in, then get the patch to the point where it can be
 committed, commit it, then move on to the notification functions, and
 probably config options after that (since that requires the context
 object to be passed much further down the chain).

 after all this is done, then i'll start looking at doing the
 cancelation stuff, which is the whole point of this exercise, if you
 remember the issue at hand ;-)

 Is there any reason not to start with the cancellation stuff, and then
 do the other stuff?  It all has to get done, sure, but at least that
 way we'd grow the new feature earlier :-).

 Your call, just pointing out the possibility.

cause it's easier to just move around existing functionality while i'm 
thinking about how to best do the cancelation stuff, and this stuff 
should get moved anyway ;-)

-garrett


---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org

Re: [PATCH] svn_client_ctx_t (was Re: Issue #1004 and Garrett Rooney)

Posted by Karl Fogel kf...@newton.ch.collab.net.
Garrett Rooney rooneg@electricjellyfish.net writes:
 ok, here's my current take on the svn_client_ctx_t issue.
 
 there's a /ton/ of stuff that we'd like to move in to this structure,
 and doing it all at once is madness, so i'm going to take an
 incremental approach, starting one at a time at the top level of the
 API (the functions in svn_client.h), and as i move more and more parts
 of the client specific data in to the context object, i'll push the
 context object further into the guts of libsvn_client and libsvn_wc.

That's the strategy I was thinking too, cool.

I don't think the client_ctx_t baton itself would ever get passed down
into libsvn_wc; but some of the things *in* the baton might be passed
as individual arguments to libsvn_wc functions, of course.

 the first thing i thought would be nice to move was the
 svn_client_auth_baton_t's that get passed to everything that accesses
 the repository directly.  i've got a first cut at that done, just
 doing so for svn_client_checkout (and the functions were affected by
 that change, progressing only far enough to make things build and pass
 make check).
 
 assuming people like this approach, i was going to finish up moving
 the auth batons in, then get the patch to the point where it can be
 committed, commit it, then move on to the notification functions, and
 probably config options after that (since that requires the context
 object to be passed much further down the chain).
 
 after all this is done, then i'll start looking at doing the
 cancelation stuff, which is the whole point of this exercise, if you
 remember the issue at hand ;-)

Is there any reason not to start with the cancellation stuff, and then
do the other stuff?  It all has to get done, sure, but at least that
way we'd grow the new feature earlier :-).

Your call, just pointing out the possibility.

(I haven't looked at the patch itself in detail yet, just due to
scheduling, but will.)

Thanks, Garrett,
-K

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org

[PATCH] svn_client_ctx_t (was Re: Issue #1004 and Garrett Rooney)

Posted by Garrett Rooney ro...@electricjellyfish.net.
ok, here's my current take on the svn_client_ctx_t issue.

there's a /ton/ of stuff that we'd like to move in to this structure, 
and doing it all at once is madness, so i'm going to take an 
incremental approach, starting one at a time at the top level of the 
API (the functions in svn_client.h), and as i move more and more parts 
of the client specific data in to the context object, i'll push the 
context object further into the guts of libsvn_client and libsvn_wc.

the first thing i thought would be nice to move was the 
svn_client_auth_baton_t's that get passed to everything that accesses 
the repository directly.  i've got a first cut at that done, just doing 
so for svn_client_checkout (and the functions were affected by that 
change, progressing only far enough to make things build and pass make 
check).

assuming people like this approach, i was going to finish up moving the 
auth batons in, then get the patch to the point where it can be 
committed, commit it, then move on to the notification functions, and 
probably config options after that (since that requires the context 
object to be passed much further down the chain).

after all this is done, then i'll start looking at doing the 
cancelation stuff, which is the whole point of this exercise, if you 
remember the issue at hand ;-)

anyway, here's the patch, and a log message.  let me know what you 
think.

-garrett

* subversion/clients/cmdline/checkout-cmd.c
* subversion/clients/cmdline/export-cmd.c
* subversion/clients/cmdline/switch-cmd.c
* subversion/clients/cmdline/switch-cmd.c
   create svn_client_ctx_t objects, stick authentication batons in them, 
and pass
   them to the underlying libsvn_client functions.

* subversion/include/svn_client.h
   (svn_client_ctx_t): new object to hold the client context.
   (svn_client_ctx_set_auth_baton, svn_client_ctx_get_auth_baton): 
functions to
    deal with caching the auth baton in the context.
   (svn_client_checkout, svn_client_update, svn_client_switch,
    svn_client_export): take a context instead of an auth_baton.

* subversion/include/svn_error_codes.h
   (SVN_ERR_CLIENT_CTX_NOT_FOUND): new error, for when the context does 
not
    contain the value we're looking for.

* subversion/libsvn_client/checkout.c
   (svn_client_checkout): take a context instead of an auth_baton, and 
grab the
    auth_baton from it when we need it.

* subversion/libsvn_client/client.h
   (svn_client__handle_externals): update prototype.

* subversion/libsvn_client/context.c
   new file, holds implementation of svn_client_ctx_*.

* subversion/libsvn_client/export.c
   (svn_client_export): take a context instead of an auth_baton, and 
grab the
    auth_baton from it when we need it.

* subversion/libsvn_client/externals.c
   (svn_client__handle_externals): take a context instead of an 
auth_baton.
   (handle_items_change_baton, handle_externals_desc_changed_baton): 
hold a
    context instead of an auth_baton.
   (all others): update calls to svn_client_update and 
svn_client_checkout to
    pass context instead of auth_baton.

* subversion/libsvn_client/switch.c
   (svn_client_switch): take an svn_client_ctx_t instead of an 
auth_baton, and
    grab the auth_baton out of it when we need it.  pass the context in 
to
    svn_client__handle_externals.

* subversion/libsvn_client/update.c
   (svn_client_update): take a context instead of an auth_baton, and 
grab the
    auth_baton out of it when we need it.  pass the context in to
    svn_client__handle_externals.


Re: Issue #1004 and Garrett Rooney

Posted by Karl Fogel kf...@newton.ch.collab.net.
Garrett Rooney rooneg@electricjellyfish.net writes:
 sure, why not...
 
 i've been looking for something to hack on now that 'the great
 doxygenification' is done ;-)

Yay!  Great.  I'll tweak the issue appropriately -- thus saving your
time for coding :-)...

-K


---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org

Re: Issue #1004 and Garrett Rooney

Posted by Garrett Rooney ro...@electricjellyfish.net.
Karl Fogel wrote:

Heh, bet that got your attention, Garrett :-).


sure did ;-)

We have a lot of issues, but also a lot of developers; I'm trying to
go through all the near issues and match them up with developers.
This is strictly voluntary, of course -- if you can't do it, just say
the word, and we'll move on to the next victim.

Anyway: would you be willing to take on issue #1004?  The summary says

   need api to interrupt long running operations via a client

...but what it's really about is creating a new svn_client_context_t
structure and passing it to all the client functions.  Greg Stein has
more comments in the issue.  Basically, this is something we've needed
for a long time; it will not only give us a way to add cancellation
and progress-meter functionality, but may also simplify the passing of
auth data, the notification functions, etc.

I remember that it's been discussed on this list, and we were all
pretty much for it, while trembling at the thought of actually
implementing it :-).  But you are brave, I know.  Well do I remember
your valour at the Caves of Arnorgaroth, and the flash of your ax as
it bit into your foe's duodenum...

Will you fight by our side again, I say?


sure, why not...

i've been looking for something to hack on now that 'the great 
doxygenification' is done ;-)

-garrett


---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org