[Caml-list] Doing compiler patch review with a dedicated mailing-list

[Caml-list] Doing compiler patch review with a dedicated mailing-list

[Adrien Nader]

Hi,

(and sorry for the mail sent a few minutes ago :) )

I'd like to know what people think about having a mailing-list for
reviews and tests of patches to the compiler and tools around it.

The idea is to do something similar to the kernel mailing-list. I mostly
like mantis and it is possible to attach files but it becomes fairly
unreadable after a while. The audience is also mostly limited to people
who are subscribed to the bug report. I hope this reduces the work and
burden of reviewers and especially commiters.

The goal is not to replace patches on mantis and you shouldn't believe
this has been blessed by the core development team (nor mentionned to
them actually). Instead, I hope this helps do quicker (and smaller?)
iteration of patches.

One example where I believe this would be useful is for the
cross-compilation patches I've started getting upstreamed around one
year ago. There are still many patchs which touch many files and
definitely need tests on platforms I don't usually run.

Another case is for patches which touch bits of the compiler almost
no-one is familiar with; I think this could help get more input.

Rules would be similar to http://linux.yyz.us/patch-format.html but less
strict (that is also mostly something to begin with).
In addition, there should be no specific reviewer or set of reviewers
for a given component; it is also known, acknowledged and perfectly fine
that the available time of reviewers varies. In practice this means you
should not refrain from commenting on a patch because someone else
usually handles a given topic.

Of course, this requires two things: a bit of infrastructure (I hear
it's much easier to create mailing-lists on ocaml.org than on inria's
servers), and people (i.e. you). Anyone interested and willing to
participate?

-- 
Adrien Nader

Re: [Caml-list] Doing compiler patch review with a dedicated mailing-list

[Simon Cruanes]

[-- Attachment #1: Type: text/plain, Size: 2626 bytes --]

Le Sat, 11 Jan 2014, Adrien Nader a écrit :
> Hi,
> 
> (and sorry for the mail sent a few minutes ago :) )
> 
> I'd like to know what people think about having a mailing-list for
> reviews and tests of patches to the compiler and tools around it.
> 
> The idea is to do something similar to the kernel mailing-list. I mostly
> like mantis and it is possible to attach files but it becomes fairly
> unreadable after a while. The audience is also mostly limited to people
> who are subscribed to the bug report. I hope this reduces the work and
> burden of reviewers and especially commiters.
> 
> The goal is not to replace patches on mantis and you shouldn't believe
> this has been blessed by the core development team (nor mentionned to
> them actually). Instead, I hope this helps do quicker (and smaller?)
> iteration of patches.
> 
> One example where I believe this would be useful is for the
> cross-compilation patches I've started getting upstreamed around one
> year ago. There are still many patchs which touch many files and
> definitely need tests on platforms I don't usually run.
> 
> Another case is for patches which touch bits of the compiler almost
> no-one is familiar with; I think this could help get more input.
> 
> Rules would be similar to http://linux.yyz.us/patch-format.html but less
> strict (that is also mostly something to begin with).
> In addition, there should be no specific reviewer or set of reviewers
> for a given component; it is also known, acknowledged and perfectly fine
> that the available time of reviewers varies. In practice this means you
> should not refrain from commenting on a patch because someone else
> usually handles a given topic.
> 
> Of course, this requires two things: a bit of infrastructure (I hear
> it's much easier to create mailing-lists on ocaml.org than on inria's
> servers), and people (i.e. you). Anyone interested and willing to
> participate?

The idea is nice. I don't know the compiler's internals and would
certainly be interested in being familiar with them. Is a new mailing
list necessary - do you envision massive traffic on the list - or could
the current ocaml list, which is usually quite quiet, be used for
compiler discussions?

The parallel with the recent suggestion to host wikis on ocaml.org to
centralize information about how to do parallelism (or other
similar topics regarding the high-level use of OCaml) may be interesting.
Would a small-ish wiki about the compiler and compiler's hacking be
useful? Can a mailing-list play the same role as a wiki?

Cheers,

-- 
Simon

[-- Attachment #2: Type: application/pgp-signature, Size: 836 bytes --]

Re: [Caml-list] Doing compiler patch review with a dedicated mailing-list

[Adrien Nader]

On Sat, Jan 11, 2014, Simon Cruanes wrote:
> Le Sat, 11 Jan 2014, Adrien Nader a écrit :
> > Hi,
> > 
> > (and sorry for the mail sent a few minutes ago :) )
> > 
> > I'd like to know what people think about having a mailing-list for
> > reviews and tests of patches to the compiler and tools around it.
> > 
> > The idea is to do something similar to the kernel mailing-list. I mostly
> > like mantis and it is possible to attach files but it becomes fairly
> > unreadable after a while. The audience is also mostly limited to people
> > who are subscribed to the bug report. I hope this reduces the work and
> > burden of reviewers and especially commiters.
> > 
> > The goal is not to replace patches on mantis and you shouldn't believe
> > this has been blessed by the core development team (nor mentionned to
> > them actually). Instead, I hope this helps do quicker (and smaller?)
> > iteration of patches.
> > 
> > One example where I believe this would be useful is for the
> > cross-compilation patches I've started getting upstreamed around one
> > year ago. There are still many patchs which touch many files and
> > definitely need tests on platforms I don't usually run.
> > 
> > Another case is for patches which touch bits of the compiler almost
> > no-one is familiar with; I think this could help get more input.
> > 
> > Rules would be similar to http://linux.yyz.us/patch-format.html but less
> > strict (that is also mostly something to begin with).
> > In addition, there should be no specific reviewer or set of reviewers
> > for a given component; it is also known, acknowledged and perfectly fine
> > that the available time of reviewers varies. In practice this means you
> > should not refrain from commenting on a patch because someone else
> > usually handles a given topic.
> > 
> > Of course, this requires two things: a bit of infrastructure (I hear
> > it's much easier to create mailing-lists on ocaml.org than on inria's
> > servers), and people (i.e. you). Anyone interested and willing to
> > participate?
> 
> The idea is nice. I don't know the compiler's internals and would
> certainly be interested in being familiar with them. Is a new mailing
> list necessary - do you envision massive traffic on the list - or could
> the current ocaml list, which is usually quite quiet, be used for
> compiler discussions?

I don't think there would be massive traffic but probably some bursts
(especially when there are 10 patches or more in a series) and that
could be annoying to some.

Also, I'd say the caml-list actually has some traffic on it.

> The parallel with the recent suggestion to host wikis on ocaml.org to
> centralize information about how to do parallelism (or other
> similar topics regarding the high-level use of OCaml) may be interesting.
> Would a small-ish wiki about the compiler and compiler's hacking be
> useful? Can a mailing-list play the same role as a wiki?

Well, I don't know compiler internals very well either so it's difficult
for me to say. However having information on how to contribute back (be
it a wiki page or a static page updated through git) would definitely be
useful as there is a growing number of people asking for such infos (at
least on IRC on #ocaml).

-- 
Adrien Nader

Re: [Caml-list] Doing compiler patch review with a dedicated mailing-list

[François Bobot]

On 13/01/2014 10:04, Adrien Nader wrote:
> On Sat, Jan 11, 2014, Simon Cruanes wrote:
>> Le Sat, 11 Jan 2014, Adrien Nader a écrit :
>>> Hi,
>>>
>>> (and sorry for the mail sent a few minutes ago :) )
>>>
>>> I'd like to know what people think about having a mailing-list for
>>> reviews and tests of patches to the compiler and tools around it.
>>>
>>> The idea is to do something similar to the kernel mailing-list. I mostly
>>> like mantis and it is possible to attach files but it becomes fairly
>>> unreadable after a while. The audience is also mostly limited to people
>>> who are subscribed to the bug report. I hope this reduces the work and
>>> burden of reviewers and especially commiters.
>>>
>>> The goal is not to replace patches on mantis and you shouldn't believe
>>> this has been blessed by the core development team (nor mentionned to
>>> them actually). Instead, I hope this helps do quicker (and smaller?)
>>> iteration of patches.
>>>

I don't know how you generate and _manage_ patches with svn. Indeed the 
linux kernel developers never used svn with their mailing-list review 
workflow and developed git for simplifying this workflow.

It seems counterproductive to have more than one place for discussing 
one thing so I think the developers must make a choice:
- keeping patch review in mantis
- going to a mailing-list review workflow and moving from svn
- going to a merge-request workflow on github, specific gitlab instance, 
bitbuckets, ...

The last two points have the benefit to allow to easily comment inside 
the patches.

The third point (at least on github) subsume the second point since you 
can answer to github issues or merge-requests by email. You can also ask 
to be notified for every issues or merge-requests of a project.

Best,

-- 
François

Re: [Caml-list] Doing compiler patch review with a dedicated mailing-list

[Gabriel Scherer]

During the last "why no ocaml on github" thread, I had a discussion
with Jonathan Protzenko, which suggested precisely that we encourage
pull requests on the existing Github mirror as a way to send patches
-- to lower the participation cost for those that are too cool to use
a bug tracker of the previous decade. Adrien's suggestion of using a
mailing-list seems equally interesting (I'm in favor of everything
that can increase community participation to OCaml), but indeed we
should probably make a choice between the several options.

The nice thing with mailing-list is that they have easy-to-browse
archives, that in my experience work more reliably than search stuff
in a github repository (in Batteries we tend to have things scattered
across a web of issues, pull-requests, and commit comments that
cross-reference each other, and I'm not always even sure where I
should write). They're also based on a stable, well-established, *free
software* stack that is there to stay (about Github, a Wise One
remarked that "yesterday the same people were commanding that we host
OCaml on Sourceforge; look where it is now!"). On the other hand,
reacting to a perceived lack of sexiness of Mantis with a
mailing-list... I'm not sure.

I'm not personally afraid of having several places where patches are
proposed (it is de facto already what happens, between the bugtracker,
direct communication/review, people that post a link to a
gitweb/github/gitlab view...), but I would be fine with blessing one
"preferred" place and documenting it where it needs be.

Re. wiki/documentation/whatever, we could take inspiration from the
fine work done in the Cambridge are (
https://github.com/ocamllabs/compiler-hacking ). Let me take this as
an occasion to remind that patches to insert enlightening comment in
the admittedly-not-always-commented-enough compiler and distribution
are welcome, and can be contributed right now.

On Mon, Jan 13, 2014 at 10:51 AM, François Bobot <francois.bobot@cea.fr> wrote:
> On 13/01/2014 10:04, Adrien Nader wrote:
>>
>> On Sat, Jan 11, 2014, Simon Cruanes wrote:
>>>
>>> Le Sat, 11 Jan 2014, Adrien Nader a écrit :
>>>>
>>>> Hi,
>>>>
>>>> (and sorry for the mail sent a few minutes ago :) )
>>>>
>>>> I'd like to know what people think about having a mailing-list for
>>>> reviews and tests of patches to the compiler and tools around it.
>>>>
>>>> The idea is to do something similar to the kernel mailing-list. I mostly
>>>> like mantis and it is possible to attach files but it becomes fairly
>>>> unreadable after a while. The audience is also mostly limited to people
>>>> who are subscribed to the bug report. I hope this reduces the work and
>>>> burden of reviewers and especially commiters.
>>>>
>>>> The goal is not to replace patches on mantis and you shouldn't believe
>>>> this has been blessed by the core development team (nor mentionned to
>>>> them actually). Instead, I hope this helps do quicker (and smaller?)
>>>> iteration of patches.
>>>>
>
> I don't know how you generate and _manage_ patches with svn. Indeed the
> linux kernel developers never used svn with their mailing-list review
> workflow and developed git for simplifying this workflow.
>
> It seems counterproductive to have more than one place for discussing one
> thing so I think the developers must make a choice:
> - keeping patch review in mantis
> - going to a mailing-list review workflow and moving from svn
> - going to a merge-request workflow on github, specific gitlab instance,
> bitbuckets, ...
>
> The last two points have the benefit to allow to easily comment inside the
> patches.
>
> The third point (at least on github) subsume the second point since you can
> answer to github issues or merge-requests by email. You can also ask to be
> notified for every issues or merge-requests of a project.
>
> Best,
>
> --
> François
>
>
> --
> Caml-list mailing list.  Subscription management and archives:
> https://sympa.inria.fr/sympa/arc/caml-list
> Beginner's list: http://groups.yahoo.com/group/ocaml_beginners
> Bug reports: http://caml.inria.fr/bin/caml-bugs

Re: [Caml-list] Doing compiler patch review with a dedicated mailing-list

[Daniel Bünzli]

Le lundi, 13 janvier 2014 à 11:27, Gabriel Scherer a écrit :
> -- to lower the participation cost for those that are too cool to use
> a bug tracker of the previous decade.  

That's a troll. What some of us are asking for is a usable bugtracker (cf. #6052). Frankly, mantis doesn't even know how to properly escape markup when you are commenting issues. Having proper ways to introduce code (and, less essential, lightweight formatting) on a bugtracker seems an essential feature to me and that's just one example, I won't even talk about having a quick look at patches that you can't even see in context but will download to your computer. If you want to use broken and inhumane software it's your problem but don't try to label those that actually prefer usable software as being attracted by cool and shiny things.

> The nice thing with mailing-list is that they have easy-to-browse
> archives,  

It's not only easy to browse, it's *great* to browse: all the web-based mailing list archives I interact with are not even able to follow a thread running across two month. I really feel we're in 2014.  
  
> (about Github, a Wise One
> remarked that "yesterday the same people were commanding that we host
> OCaml on Sourceforge; look where it is now!").  

Sure, if you move to that kind of (free or paid) service always factor in the costs of entering and leaving it, as it certainly won't help you forever. What's forever anyways ?  
  
> I'm not personally afraid of having several places where patches are
> proposed

I think that's a rather bad idea.  

Daniel  

P.S. Since you mentioned gitlab et al, I was recently pointed to this aswell: http://phabricator.org/ never interacted with though.

Re: [Caml-list] Doing compiler patch review with a dedicated mailing-list

[Gabriel Scherer]

On Mon, Jan 13, 2014 at 12:14 PM, Daniel Bünzli
<daniel.buenzli@erratique.ch> wrote:
> It's not only easy to browse, it's *great* to browse: all the web-based mailing list archives I interact with are not even able to follow a thread running across two month. I really feel we're in 2014.

This is a bit too snarky for me to guess what you want.

I'm already doing a few reviews on mantis, and occasionally uses the
github in-patch-commenting interface when people send a link to a
github-hosted patch (eg. http://caml.inria.fr/mantis/view.php?id=6274
). I would be ready to experiment with a mailing-list or a specialized
review tool, based on what people would prefer. We have a solid github
mirror in place ( github.com/ocaml/ocaml ), people of OCamllabs have
kindly offered to host mailing-lists on several occasions, and have
experimented with, for example, gitlab in the past.

People of the list, if you have sent or reviewed patches or consider
doing so in the future, do you have a strong preference? If your
peer/friend/colleague wishes to contribute and asks you the place to
go, what will you wish we had?

(Keep in mind that we don't have a lot of patches sent around for now
(less than a dozen a month), and even less people*time to do the
reviews, so we don't necessarily need a highly tuned über-process.)

On Mon, Jan 13, 2014 at 12:14 PM, Daniel Bünzli
<daniel.buenzli@erratique.ch> wrote:
> Le lundi, 13 janvier 2014 à 11:27, Gabriel Scherer a écrit :
>> -- to lower the participation cost for those that are too cool to use
>> a bug tracker of the previous decade.
>
> That's a troll. What some of us are asking for is a usable bugtracker (cf. #6052). Frankly, mantis doesn't even know how to properly escape markup when you are commenting issues. Having proper ways to introduce code (and, less essential, lightweight formatting) on a bugtracker seems an essential feature to me and that's just one example, I won't even talk about having a quick look at patches that you can't even see in context but will download to your computer. If you want to use broken and inhumane software it's your problem but don't try to label those that actually prefer usable software as being attracted by cool and shiny things.
>
>> The nice thing with mailing-list is that they have easy-to-browse
>> archives,
>
> It's not only easy to browse, it's *great* to browse: all the web-based mailing list archives I interact with are not even able to follow a thread running across two month. I really feel we're in 2014.
>
>> (about Github, a Wise One
>> remarked that "yesterday the same people were commanding that we host
>> OCaml on Sourceforge; look where it is now!").
>
> Sure, if you move to that kind of (free or paid) service always factor in the costs of entering and leaving it, as it certainly won't help you forever. What's forever anyways ?
>
>> I'm not personally afraid of having several places where patches are
>> proposed
>
> I think that's a rather bad idea.
>
> Daniel
>
> P.S. Since you mentioned gitlab et al, I was recently pointed to this aswell: http://phabricator.org/ never interacted with though.

Re: [Caml-list] Doing compiler patch review with a dedicated mailing-list

[Thomas Refis]

2014/1/13 Gabriel Scherer <gabriel.scherer@gmail.com>:
> On Mon, Jan 13, 2014 at 12:14 PM, Daniel Bünzli
> <daniel.buenzli@erratique.ch> wrote:
>> It's not only easy to browse, it's *great* to browse: all the web-based mailing list archives I interact with are not even able to follow a thread running across two month. I really feel we're in 2014.
>
> This is a bit too snarky for me to guess what you want.

I'd have thought that part was pretty obvious. If the discussion
starts on the 28th of a month, and is continued the next month. The
"sympa" interface gives you no easy way to follow the thread. You need
to "select" the second month, and find the thread again.

> I'm already doing a few reviews on mantis, and occasionally uses the
> github in-patch-commenting interface when people send a link to a
> github-hosted patch (eg. http://caml.inria.fr/mantis/view.php?id=6274
> ).

Since we're talking about it: I was a bit confused with that review.
Although *I* find the github interface nicer than mantis and the "in
patch comments" really useful, I didn't know whether I should answer
your comments on github or on mantis, and people on mantis might not
see the comments you did.
So I'm not sure that's really a good approach. (But of course that
particular patch was about ocamldoc, which no one really cares about,
so I guess that's ok)

Thomas.

Re: [Caml-list] Doing compiler patch review with a dedicated mailing-list

[Gabriel Scherer]

> I'd have thought that part was pretty obvious.

The remark was clear, but I am more interested in what Daniel *would
like* to see used. He is quite good at ranting against just about
everything, so I have no indication that he would be merrier with
something else.

On Mon, Jan 13, 2014 at 2:43 PM, Thomas Refis <thomas.refis@gmail.com> wrote:
> 2014/1/13 Gabriel Scherer <gabriel.scherer@gmail.com>:
>> On Mon, Jan 13, 2014 at 12:14 PM, Daniel Bünzli
>> <daniel.buenzli@erratique.ch> wrote:
>>> It's not only easy to browse, it's *great* to browse: all the web-based mailing list archives I interact with are not even able to follow a thread running across two month. I really feel we're in 2014.
>>
>> This is a bit too snarky for me to guess what you want.
>
> I'd have thought that part was pretty obvious. If the discussion
> starts on the 28th of a month, and is continued the next month. The
> "sympa" interface gives you no easy way to follow the thread. You need
> to "select" the second month, and find the thread again.
>
>> I'm already doing a few reviews on mantis, and occasionally uses the
>> github in-patch-commenting interface when people send a link to a
>> github-hosted patch (eg. http://caml.inria.fr/mantis/view.php?id=6274
>> ).
>
> Since we're talking about it: I was a bit confused with that review.
> Although *I* find the github interface nicer than mantis and the "in
> patch comments" really useful, I didn't know whether I should answer
> your comments on github or on mantis, and people on mantis might not
> see the comments you did.
> So I'm not sure that's really a good approach. (But of course that
> particular patch was about ocamldoc, which no one really cares about,
> so I guess that's ok)
>
> Thomas.

Re: [Caml-list] Doing compiler patch review with a dedicated mailing-list

[Daniel Bünzli]

Le lundi, 13 janvier 2014 à 14:26, Gabriel Scherer a écrit :

> On Mon, Jan 13, 2014 at 12:14 PM, Daniel Bünzli
> <daniel.buenzli@erratique.ch (mailto:daniel.buenzli@erratique.ch)> wrote:
> > It's not only easy to browse, it's *great* to browse: all the web-based mailing list archives I interact with are not even able to follow a thread running across two month. I really feel we're in 2014.
>  
> This is a bit too snarky for me to guess what you want.

Regarding what I want, I think #6052 is pretty clear, as is its resolution, which I perfectly understand: it's up to the dev team to choose the tools and workflows that suit them. But what I wanted to say is: if you are going to change your practice, at least, don't replace it by something that seems more broken (at least to me) than what we currently have.

Best,

Daniel

Re: [Caml-list] Doing compiler patch review with a dedicated mailing-list

[Simon Cruanes]

[-- Attachment #1: Type: text/plain, Size: 1836 bytes --]

Le Mon, 13 Jan 2014, Thomas Refis a écrit :
> 2014/1/13 Gabriel Scherer <gabriel.scherer@gmail.com>:
> > On Mon, Jan 13, 2014 at 12:14 PM, Daniel Bünzli
> > <daniel.buenzli@erratique.ch> wrote:
> >> It's not only easy to browse, it's *great* to browse: all the web-based mailing list archives I interact with are not even able to follow a thread running across two month. I really feel we're in 2014.
> >
> > This is a bit too snarky for me to guess what you want.
> 
> I'd have thought that part was pretty obvious. If the discussion
> starts on the 28th of a month, and is continued the next month. The
> "sympa" interface gives you no easy way to follow the thread. You need
> to "select" the second month, and find the thread again.
> 
> > I'm already doing a few reviews on mantis, and occasionally uses the
> > github in-patch-commenting interface when people send a link to a
> > github-hosted patch (eg. http://caml.inria.fr/mantis/view.php?id=6274
> > ).
> 
> Since we're talking about it: I was a bit confused with that review.
> Although *I* find the github interface nicer than mantis and the "in
> patch comments" really useful, I didn't know whether I should answer
> your comments on github or on mantis, and people on mantis might not
> see the comments you did.
> So I'm not sure that's really a good approach. (But of course that
> particular patch was about ocamldoc, which no one really cares about,
> so I guess that's ok)

I like the github interface, but it's clear that it also comes with
vendor lock-in. To mitigate it, may it be possible to duplicate every
mail from github (issues, pull requests, i.e. any interesting
discussion) to a separate mailing-list that would serve as
archive+backup? The question is "how to store the mails for archive".

my 2 cents

-- 
Simon

[-- Attachment #2: Type: application/pgp-signature, Size: 819 bytes --]

Re: [Caml-list] Doing compiler patch review with a dedicated mailing-list

[Kakadu]

Qt Gerrit [1] is the most fascinating review system I have ever seen.
Everybody can push commits and add reviewers for it. CI is enabled for
every change. Any chance that OCaml will  have the same available?

[1] https://codereview.qt-project.org/#change,75024

Cheers,
Kakadu

On Mon, Jan 13, 2014 at 1:43 PM, Thomas Refis <thomas.refis@gmail.com> wrote:
> 2014/1/13 Gabriel Scherer <gabriel.scherer@gmail.com>:
>> On Mon, Jan 13, 2014 at 12:14 PM, Daniel Bünzli
>> <daniel.buenzli@erratique.ch> wrote:
>>> It's not only easy to browse, it's *great* to browse: all the web-based mailing list archives I interact with are not even able to follow a thread running across two month. I really feel we're in 2014.
>>
>> This is a bit too snarky for me to guess what you want.
>
> I'd have thought that part was pretty obvious. If the discussion
> starts on the 28th of a month, and is continued the next month. The
> "sympa" interface gives you no easy way to follow the thread. You need
> to "select" the second month, and find the thread again.
>
>> I'm already doing a few reviews on mantis, and occasionally uses the
>> github in-patch-commenting interface when people send a link to a
>> github-hosted patch (eg. http://caml.inria.fr/mantis/view.php?id=6274
>> ).
>
> Since we're talking about it: I was a bit confused with that review.
> Although *I* find the github interface nicer than mantis and the "in
> patch comments" really useful, I didn't know whether I should answer
> your comments on github or on mantis, and people on mantis might not
> see the comments you did.
> So I'm not sure that's really a good approach. (But of course that
> particular patch was about ocamldoc, which no one really cares about,
> so I guess that's ok)
>
> Thomas.
>
> --
> Caml-list mailing list.  Subscription management and archives:
> https://sympa.inria.fr/sympa/arc/caml-list
> Beginner's list: http://groups.yahoo.com/group/ocaml_beginners
> Bug reports: http://caml.inria.fr/bin/caml-bugs

Re: [Caml-list] Doing compiler patch review with a dedicated mailing-list

[Török Edwin]

On 01/13/2014 03:57 PM, Simon Cruanes wrote:
> Le Mon, 13 Jan 2014, Thomas Refis a écrit :
>>
>>> I'm already doing a few reviews on mantis, and occasionally uses the
>>> github in-patch-commenting interface when people send a link to a
>>> github-hosted patch (eg. http://caml.inria.fr/mantis/view.php?id=6274
>>> ).
>>
>> Since we're talking about it: I was a bit confused with that review.
>> Although *I* find the github interface nicer than mantis and the "in
>> patch comments" really useful, I didn't know whether I should answer
>> your comments on github or on mantis, and people on mantis might not
>> see the comments you did.
>> So I'm not sure that's really a good approach. (But of course that
>> particular patch was about ocamldoc, which no one really cares about,
>> so I guess that's ok)
> 
> I like the github interface, but it's clear that it also comes with
> vendor lock-in.

It is probably a matter of what you're used with, for me - coming from Bugzilla - reporting an issue on Github is more complicated,
and less obvious than on Bugzilla.
For example I don't know how to mark an issue as a 'feature request', or 'enhancement' as there are no severities; and I always
have to be careful of special characters that mess up formatting in the body of the issue, and need to put those in <pre> blocks.

I didn't have any considerable problems reporting bugs on Mantis, in fact its probably easier for me than Github, although
I have to admit that I didn't really try searching for bugs on Mantis or Github, and Github might have an advantage there.

Best regards,
--Edwin

Re: [Caml-list] Doing compiler patch review with a dedicated mailing-list

[Yotam Barnoy]

[-- Attachment #1: Type: text/plain, Size: 5984 bytes --]

On Mon, Jan 13, 2014 at 5:27 AM, Gabriel Scherer
<gabriel.scherer@gmail.com>wrote:

> (about Github, a Wise One
> remarked that "yesterday the same people were commanding that we host
> OCaml on Sourceforge; look where it is now!").
>

I'm just going to focus on this comment: I don't see anything wrong with
this. Companies find good developers by using recruiters and headhunters.
Open source projects do so by, among other things, going to the most
popular places for developers to hang out. Back when sourceforge was
popular, it was a great place for people looking to get involved in open
source projects (like myself). The proper view on this in my opinion is,
how many great devs were *not* introduced to ocaml and did *not* join ocaml
development because ocaml wasn't where the people were? Is it worth the
little bit of hassle to get long-time devs to switch their methodologies
for the benefit of getting more developers into the project/language? In my
humble opinion, the answer is absolutely yes.

Regardless, sourceforge's other advantages in its best days were minimal
for an organization that can host its own repository and issue tracker.
Github has a much bigger advantage in that it comes with a whole ecosystem
of tools that make development that much easier. Github is also a much
bigger player at a time when open source software is a much bigger player
in the world: for example, employees are looking to add open source
projects to their resumes, and they're looking mostly on github (Btw I
realize that ocaml has a mirror on github, but consider that a read-only
mirror is not a lively, encouraging environment with which to interact).

My main problem with Mantis is that one misses out on the ability to see
patches automatically applied to code, and to comment on specific parts of
patches. These things are huge productivity boosts and make it much easier
for many people to participate and comment on patches. Also, being able to
fork and then just post a pull request means that I don't have to mess with
managing patch files at all. There's a good reason the world is moving in
this direction.

I understand the trepidation to have all this extra value trapped within
github, but I have a strong hunch that if github starts losing to another,
better competitor, bigger projects will move out first, and tools will be
developed using the github api to extract and duplicate the metadata. In
fact, a quick search turned up several of these kinds of things already,
just because they're easy to build, e.g.
https://github.com/joeyh/github-backup

It's also useful to see how other languages do things. Mono, Scala and
Clojure are all on github. Of the three, Mono and Clojure don't use the
'issues' feature of github, but all three use pull requests (and the rich
features they enable). Clojure uses Jira (which looks extremely powerful
and is free to open source projects) to manage its issues and planning.
Mono uses bugzilla for managing issues, but also allows/uses pull requests.

Regarding the mailing list, my experience from another open source project
suggests that mailing lists are better for discussion of bigger features,
or for pointing out specific patches for review, rather than spreading out
the discussion in multiple places, but I could definitely be wrong about
this.

-Yotam


On Mon, Jan 13, 2014 at 10:51 AM, François Bobot <francois.bobot@cea.fr>
> wrote:
> > On 13/01/2014 10:04, Adrien Nader wrote:
> >>
> >> On Sat, Jan 11, 2014, Simon Cruanes wrote:
> >>>
> >>> Le Sat, 11 Jan 2014, Adrien Nader a écrit :
> >>>>
> >>>> Hi,
> >>>>
> >>>> (and sorry for the mail sent a few minutes ago :) )
> >>>>
> >>>> I'd like to know what people think about having a mailing-list for
> >>>> reviews and tests of patches to the compiler and tools around it.
> >>>>
> >>>> The idea is to do something similar to the kernel mailing-list. I
> mostly
> >>>> like mantis and it is possible to attach files but it becomes fairly
> >>>> unreadable after a while. The audience is also mostly limited to
> people
> >>>> who are subscribed to the bug report. I hope this reduces the work and
> >>>> burden of reviewers and especially commiters.
> >>>>
> >>>> The goal is not to replace patches on mantis and you shouldn't believe
> >>>> this has been blessed by the core development team (nor mentionned to
> >>>> them actually). Instead, I hope this helps do quicker (and smaller?)
> >>>> iteration of patches.
> >>>>
> >
> > I don't know how you generate and _manage_ patches with svn. Indeed the
> > linux kernel developers never used svn with their mailing-list review
> > workflow and developed git for simplifying this workflow.
> >
> > It seems counterproductive to have more than one place for discussing one
> > thing so I think the developers must make a choice:
> > - keeping patch review in mantis
> > - going to a mailing-list review workflow and moving from svn
> > - going to a merge-request workflow on github, specific gitlab instance,
> > bitbuckets, ...
> >
> > The last two points have the benefit to allow to easily comment inside
> the
> > patches.
> >
> > The third point (at least on github) subsume the second point since you
> can
> > answer to github issues or merge-requests by email. You can also ask to
> be
> > notified for every issues or merge-requests of a project.
> >
> > Best,
> >
> > --
> > François
> >
> >
> > --
> > Caml-list mailing list.  Subscription management and archives:
> > https://sympa.inria.fr/sympa/arc/caml-list
> > Beginner's list: http://groups.yahoo.com/group/ocaml_beginners
> > Bug reports: http://caml.inria.fr/bin/caml-bugs
>
> --
> Caml-list mailing list.  Subscription management and archives:
> https://sympa.inria.fr/sympa/arc/caml-list
> Beginner's list: http://groups.yahoo.com/group/ocaml_beginners
> Bug reports: http://caml.inria.fr/bin/caml-bugs
>

[-- Attachment #2: Type: text/html, Size: 7666 bytes --]

Re: [Caml-list] Doing compiler patch review with a dedicated mailing-list

[Adrien Nader]

On Mon, Jan 13, 2014, Gabriel Scherer wrote:
> On Mon, Jan 13, 2014 at 12:14 PM, Daniel Bünzli
> <daniel.buenzli@erratique.ch> wrote:
> > It's not only easy to browse, it's *great* to browse: all the web-based mailing list archives I interact with are not even able to follow a thread running across two month. I really feel we're in 2014.
> 
> This is a bit too snarky for me to guess what you want.
> 
> I'm already doing a few reviews on mantis, and occasionally uses the
> github in-patch-commenting interface when people send a link to a
> github-hosted patch (eg. http://caml.inria.fr/mantis/view.php?id=6274
> ). I would be ready to experiment with a mailing-list or a specialized
> review tool, based on what people would prefer. We have a solid github
> mirror in place ( github.com/ocaml/ocaml ), people of OCamllabs have
> kindly offered to host mailing-lists on several occasions, and have
> experimented with, for example, gitlab in the past.
> 
> People of the list, if you have sent or reviewed patches or consider
> doing so in the future, do you have a strong preference? If your
> peer/friend/colleague wishes to contribute and asks you the place to
> go, what will you wish we had?
> 
> (Keep in mind that we don't have a lot of patches sent around for now
> (less than a dozen a month), and even less people*time to do the
> reviews, so we don't necessarily need a highly tuned über-process.)

Reducing reviewer work and/or having more reviewers was the actual goal
behind my proposal.
Another core aspect is that setting up a mailing-list and putting up a
"howto contribute" page takes a few minutes of work unlike migrations of
whole systems or the use of forges(*).

I was also not thinking of replacing patches and reviews on mantis, more
having an additional tool to conduct code review although, of course, I
would prefer a single place for that. In any case, having something
agreed upon and documented is probably what matters most.

On a personal note, I also prefer getting fed with emails for which I
will use a decent client with an editor I enjoy rather than a web
browser (especially when there's flash on the pages).

(*) A word about phabricator too: it has some very nice things on paper
and on demo but I've found it makes things easier for frequent
users/devs and very bad for others

-- 
Adrien Nader

Re: [Caml-list] Doing compiler patch review with a dedicated mailing-list

[Simon Cruanes]

[-- Attachment #1: Type: text/plain, Size: 1006 bytes --]

Le Mon, 13 Jan 2014, Adrien Nader a écrit :

> Reducing reviewer work and/or having more reviewers was the actual goal
> behind my proposal.
> Another core aspect is that setting up a mailing-list and putting up a
> "howto contribute" page takes a few minutes of work unlike migrations of
> whole systems or the use of forges(*).
> 
> I was also not thinking of replacing patches and reviews on mantis, more
> having an additional tool to conduct code review although, of course, I
> would prefer a single place for that. In any case, having something
> agreed upon and documented is probably what matters most.
> 
> On a personal note, I also prefer getting fed with emails for which I
> will use a decent client with an editor I enjoy rather than a web
> browser (especially when there's flash on the pages).

So, what's actually needed and sufficient, you believe, is a proper
mailing list manager, which provides a friendly interface to navigate
through the archive?

-- 
Simon

[-- Attachment #2: Type: application/pgp-signature, Size: 836 bytes --]

Re: [Caml-list] Doing compiler patch review with a dedicated mailing-list

[Adrien Nader]

On Mon, Jan 13, 2014, Simon Cruanes wrote:
> Le Mon, 13 Jan 2014, Adrien Nader a écrit :
> 
> > Reducing reviewer work and/or having more reviewers was the actual goal
> > behind my proposal.
> > Another core aspect is that setting up a mailing-list and putting up a
> > "howto contribute" page takes a few minutes of work unlike migrations of
> > whole systems or the use of forges(*).
> > 
> > I was also not thinking of replacing patches and reviews on mantis, more
> > having an additional tool to conduct code review although, of course, I
> > would prefer a single place for that. In any case, having something
> > agreed upon and documented is probably what matters most.
> > 
> > On a personal note, I also prefer getting fed with emails for which I
> > will use a decent client with an editor I enjoy rather than a web
> > browser (especially when there's flash on the pages).
> 
> So, what's actually needed and sufficient, you believe, is a proper
> mailing list manager, which provides a friendly interface to navigate
> through the archive?

Well, I believe a mailing-list would be suited to code review. I don't
even mind that archives require a minor hoop to navigate in long
messages since that wouldn't be the main way it is used. I can't tell if
it's needed or sufficient though.

-- 
Adrien Nader

Re: [Caml-list] Doing compiler patch review with a dedicated mailing-list

[Gabriel Kerneis]

On Mon, Jan 13, 2014 at 02:26:29PM +0100, Gabriel Scherer wrote:
> People of the list, if you have sent or reviewed patches or consider
> doing so in the future, do you have a strong preference?

1/ git-send-email (strong preference)
2/ github (weak preference)
3/ mantis (strong aversion)

-- 
Gabriel

Re: [Caml-list] Doing compiler patch review with a dedicated mailing-list

[François Bobot]

On Mon, Jan 13, 2014 at 02:26:29PM +0100, Gabriel Scherer wrote:
> People of the list, if you have sent or reviewed patches or consider
> doing so in the future, do you have a strong preference?

1/ github (strong preference)
2/ gitlab (weak preference)
3/ git-send-email (never used)
4/ mantis (strong aversion)

-- 
François

Re: [Caml-list] Doing compiler patch review with a dedicated mailing-list

[Thomas Gazagnaire]

> 1/ github (strong preference)
> 2/ gitlab (weak preference)
> 3/ git-send-email (never used)
> 4/ mantis (strong aversion)

Same here.

Thomas

Re: [Caml-list] Doing compiler patch review with a dedicated mailing-list

[Markus Mottl]

On Tue, Jan 14, 2014 at 8:27 AM, Thomas Gazagnaire
<thomas@gazagnaire.org> wrote:
>> 1/ github (strong preference)
>> 2/ gitlab (weak preference)
>> 3/ git-send-email (never used)
>> 4/ mantis (strong aversion)
>
> Same here.

And here.

Regards,
Markus

-- 
Markus Mottl        http://www.ocaml.info        markus.mottl@gmail.com

Re: [Caml-list] Doing compiler patch review with a dedicated mailing-list

[Simon Cruanes]

[-- Attachment #1: Type: text/plain, Size: 469 bytes --]

Le Tue, 14 Jan 2014, Markus Mottl a écrit :

> On Tue, Jan 14, 2014 at 8:27 AM, Thomas Gazagnaire
> <thomas@gazagnaire.org> wrote:
> >> 1/ github (strong preference)
> >> 2/ gitlab (weak preference)
> >> 3/ git-send-email (never used)
> >> 4/ mantis (strong aversion)
> >
> > Same here.
> 
> And here.

Same, but I don't know gitlab so

1/ github (preference)
2/ git-send-mail (why not, never used but I like git)
3/ mantis (reluctant)

-- 
Simon

[-- Attachment #2: Type: application/pgp-signature, Size: 819 bytes --]

Re: [Caml-list] Doing compiler patch review with a dedicated mailing-list

[Amir Chaudhry]

Hi folks,

In an attempt to be more systematic I've made a public Google form. I've added the last 5 responses into it (Simon, Markus, Thomas, Francois, Gabriel) so if there are others who'd like to note their preferences but perhaps didn't want to mail the list, they can now do so.

If anyone thinks there should be more options or would like access to the form itself, please email me directly.  

Form: https://docs.google.com/forms/d/1QWhqJRv1yPvdi6E3AiqbvUwlqGorV_Wbk7h_JYuDUiQ/viewform

Results: https://docs.google.com/forms/d/1QWhqJRv1yPvdi6E3AiqbvUwlqGorV_Wbk7h_JYuDUiQ/viewanalytics

I hope you find this useful,
Amir

On 14 Jan 2014, at 14:12, Simon Cruanes <simon.cruanes.2007@m4x.org> wrote:

> Le Tue, 14 Jan 2014, Markus Mottl a écrit :
> 
>> On Tue, Jan 14, 2014 at 8:27 AM, Thomas Gazagnaire
>> <thomas@gazagnaire.org> wrote:
>>>> 1/ github (strong preference)
>>>> 2/ gitlab (weak preference)
>>>> 3/ git-send-email (never used)
>>>> 4/ mantis (strong aversion)
>>> 
>>> Same here.
>> 
>> And here.
> 
> Same, but I don't know gitlab so
> 
> 1/ github (preference)
> 2/ git-send-mail (why not, never used but I like git)
> 3/ mantis (reluctant)
> 
> -- 
> Simon

Re: [Caml-list] Doing compiler patch review with a dedicated mailing-list

[François Bobot]

> Same, but I don't know gitlab so

Just for information

http://gitlab.org/

GitLab is an open source private github. Just less mature and 
feature-full than github but it is open source, very alive and on your 
server. Very useful for developer teams which can't use a public github.

Best,

-- 
François

Re: [Caml-list] Doing compiler patch review with a dedicated mailing-list

[Anil Madhavapeddy]

On 14 Jan 2014, at 15:09, François Bobot <francois.bobot@cea.fr> wrote:

>> Same, but I don't know gitlab so
> 
> Just for information
> 
> http://gitlab.org/
> 
> GitLab is an open source private github. Just less mature and feature-full than github but it is open source, very alive and on your server. Very useful for developer teams which can't use a public github.

I installed an experimental read-only version of GitLab at:
http://git.ocaml.org/ocaml-compiler/ocaml/tree/trunk

I'm not in favour of maintaining it on an ongoing basis though.  It's a bit of a Ruby hydra to keep updated and secure...

-anil

Re: [Caml-list] Doing compiler patch review with a dedicated mailing-list

[Richard W.M. Jones]

On Mon, Jan 13, 2014 at 01:58:11PM +0000, Kakadu wrote:
> Qt Gerrit [1] is the most fascinating review system I have ever seen.
> Everybody can push commits and add reviewers for it. CI is enabled for
> every change. Any chance that OCaml will  have the same available?
> 
> [1] https://codereview.qt-project.org/#change,75024

Late to the party as usual, but I use Gerrit for Openstack, and
it's horrible.

Mailing lists *much* preferred for patch reviews.

Rich.

-- 
Richard Jones
Red Hat