(1) get rid of there being more than 1 canonical place to
go!
That won’t help for this issue, just wanted to say ATS is not an exception.
I don’t understand neither why so much stuff seems to be hosted on both
SourceForge and Github or on both SourceForge and BitBucket (add to that,
the content versioning systems issue, all being incompatible with each
other… another story 
)
if the easiest thing is to simply use SF for a single tar.gz or
whatever then go with that. but get rid of whatever else is on SF
Essentially, it offers two versions:
sourceforge-stable (released)
sourceforge-unstable (git-repo)
Whenever I add a new feature, I often update the git-repo accordingly.
When teaching, I simply tell people to do a git-pull and re-compile. It is
very convenient.
Note that generating the bootstrapping C code for ATS2 requires the use
ATS1. I don’t really want an average user to be exposed to ATS1. However, a
developer does need ATS1.
In the long run, I don’t think either sourceforge or github is ideal for
hosting ATS.
One needs to build a web-based system to do it, offering features like
setting up tutorials.
If there is a will, then there shall be a way :)On Thursday, August 7, 2014 6:53:25 PM UTC-4, Raoul Duke wrote:
ah. their faq sucks. at first reading it looks like you can go over.
but then you can’t so who knows. jerks. dumb jerks. make it clear!
sheesh! (see? github sucks, too.) stackoverflow says you can
transiently go over, but no consistently.i don’t understand enough here. are the generated C sources important?
why can’t they “just” be re-gene’d locally by each person? or why
can’t they be tar gzip -9’d and saved in the repo? etc. etc. etc.if the easiest thing is to simply use SF for a single tar.gz or
whatever then go with that. but get rid of whatever else is on SF. or,
it is “easy” toget 5GB free from major services
(http://goo.gl/YnzZ86).probably i still don’t get it. i want there to be a way to make it all
really simple and almost simplistic :-}
in other words: the project lacks a stable vs. dev branch?
Sorry for not being clear.
If one wants to use ATS2, then please get the released tar.gz
at sourceforge.
For the few ones what would like to help develop ATS2, please use
the github version.
I have already merged your patch (submitted to sourceforge/ats2-lang)
into the git-repo for ATS2. I was trying to say that the merge could have
been done directly if your patch were made for the github-version. In
any case, it is not a big deal for me to take a few extra steps to do
the merge.On Thursday, August 7, 2014 12:58:50 AM UTC-4, Raoul Duke wrote:
i think somehow i’m failing to communicate. i do not understand how
this isn’t a big deal if you actually want people to use the system.
having directions that are actually correct is i should think
important?i tried to submit a fork/patch request for the INSTALL file in git on
sourceforge, (i actually want things to be good here!) but that was
based on my targz experience, and now the experience using git is
different and nothing anywhere seems to correctly clearly document how
to make it work. so my efforts seem to be for naught already. dogged.i mean i would have had just as much trouble and eventual success
without reading the INSTALL file. which seems unfortunate.i know ats isn’t a commercial production thing, but i really think if
you want to make sure nothing stands in the way of people trying it
out. we’ve got enough computer troubles day in day out. activation
energy to put up with bumps in the road can vary, but i suspect you
are already making some otherwise interested people give up.i mean in this day and age if it doesn’t install and run in 5 minutes
i pretty much write things off. i mean, if the install isn’t fun how
can the rest of it be?$0.02
Yes, as make reminds you if you blindly followed the the doc . I was
surprised too, by the steps suggested in the doc, as usually, one is
supposed to run configure (and to not forget the important
--prefix=<path> option) before make / make install.Le jeudi 7 août 2014 04:52:39 UTC+2, Raoul Duke a écrit :
i think the instructions are missing the ./configure step? the
makefile tells me, but the instructions should say it as well, then.
A set of problems I just ran into:
- the repo at GitHub - githwxi/ATS-Postiats: ATS2: Unleashing the Potentials of Types and Templates is missing an
autogen.sh- when I ran through the steps manually, ./configure was generated but
running it did not generate a makefile- Makefile_devl seems to depend on an existing ats install:
Professor Xi may tell more, I just can say I remember he said: so far,
generating ATS2 C sources requires ATS1. Anyway, you will need an ATS
compiler installed to compile to C then compile C to binary. The original
sources are not C sources, they are ATS1 sources (and will be ATS2 sources
in the future).
It does, though in an unusual way:
Stable releases are at sourceforge
Development is happening at github.On Thursday, August 7, 2014 2:43:17 PM UTC-4, Raoul Duke wrote:
in other words: the project lacks a stable vs. dev branch?
I don’t think we necessarily need to follow the recipe of ./configure &&
make && make install
However, I do think it makes sense from a user standpoint to have a uniform
build process; as Duke has suggested, it does make it easier to contribute
to the github repo. I can work from the github (Makefile_devl) and try to
integrate any additional functionality (probably mainly the install phase).
The only caveat I can think of is that most people probably
wouldn’t/shouldn’t do ‘make install’ on a git version, but I think that’s a
much smaller problem to deal with at the moment. Perhaps a simple test can
be done to check for ‘.git’.On Thursday, August 7, 2014 2:45:36 AM UTC-4, Yannick Duchêne wrote:
Yes, as
makereminds you if you blindly followed the the doc
surprised too, by the steps suggested in the doc, as usually, one is
supposed to run configure (and to not forget the important
--prefix=<path>option) before make / make install.Le jeudi 7 août 2014 04:52:39 UTC+2, Raoul Duke a écrit :
i think the instructions are missing the ./configure step? the
makefile tells me, but the instructions should say it as well, then.
problem #2: gmp.h isn’t mentioned as a pre-req that i’ve seen?
In file included from pats_intinf_dats.c:41:0:
libc/CATS/gmp.cats:40:17: fatal error: gmp.h: No such file or directory
#include <gmp.h>
^
compilation terminated.
gcc -O2 -I. -I./ccomp/runtime -c -o pats_global_dats.o pats_global_dats.c
gcc -O2 -I. -I./ccomp/runtime -c -o pats_basics_dats.o pats_basics_dats.c
gcc -O2 -I. -I./ccomp/runtime -c -o pats_stamp_dats.o pats_stamp_dats.c
gcc -O2 -I. -I./ccomp/runtime -c -o pats_symbol_dats.o pats_symbol_dats.c
make[1]: *** [pats_intinf_dats.o] Error 1
make[1]: *** Waiting for unfinished jobs…
make[1]: Leaving directory
`/home/superman/Downloads/tmp/ATS2-Postiats-0.1.1/src/CBOOT’
make: *** [src2_patsopt] Error 2
Interesting … what is on bit-bucked? I know another user had some code on
bitbucket.
Brandon Barker
brandon…@gmail.comOn Thu, Aug 7, 2014 at 2:39 PM, ‘Yannick Duchêne’ via ats-lang-users < ats-lan...@googlegroups.com> wrote:
Le jeudi 7 août 2014 19:40:00 UTC+2, Raoul Duke a écrit :
(1) get rid of there being more than 1 canonical place to
go!That won’t help for this issue, just wanted to say ATS is not an
exception. I don’t understand neither why so much stuff seems to be hosted
on both SourceForge and Github or on both SourceForge and BitBucket (add to
that, the content versioning systems issue, all being incompatible with
each other… another story–
You received this message because you are subscribed to the Google Groups
“ats-lang-users” group.
To unsubscribe from this group and stop receiving emails from it, send an
email to ats-lang-user...@googlegroups.com.
To post to this group, send email to ats-lan...@googlegroups.com.
Visit this group at http://groups.google.com/group/ats-lang-users.
To view this discussion on the web visit
https://groups.google.com/d/msgid/ats-lang-users/e6218acb-a0be-4ccf-bfc4-829d13811b25%40googlegroups.com
https://groups.google.com/d/msgid/ats-lang-users/e6218acb-a0be-4ccf-bfc4-829d13811b25%40googlegroups.com?utm_medium=email&utm_source=footer
.
hi,On Thu, Aug 7, 2014 at 2:53 PM, gmhwxi gmh...@gmail.com wrote:
Please note that the sourceforge-version essentially
consists of C code generated from the github-version.
To me, it is confusing to put these two versions together.The released version is sourceforge-stable and the git-repo
at sourceforge is sourceforge-unstable.It is possible to have github-stable/github-unstable in the future,
but that would be a different story. In my opinion, no user should
be really using the gitbub-version unless he/she also wants
to help develop ATS2.
all the more reason to put it all in one service.
you can have more than 1 project on github, with really clear names
and really clear readme.md files that explain succinctly the
relationship.
rather than it having to come up in mailing lists.
(i still do not really even fully understand what you said above. i
still really have no finalized correct idea of where i’m supposed to
go to get things. or where i’m supposed to go to submit patches. or
how they are related to what other new people are going to try to
download and read and install. etc. this flow might make sense to you,
but i think it is a stretch to say it is a sustainable thing that will
help grow an ecosystem for ats.)
If someone were to do it:
Would a straight port be relatively easy?
Would a s straight port be advisable?
(Not that I intend to do so anytime soon, but it might be good to
record possible ideas for posterity)
Brandon Barker
brandon…@gmail.comOn Thu, Aug 7, 2014 at 4:40 PM, gmhwxi gmh...@gmail.com wrote:
Correct in general.
and will be ATS2 sources in the future
I have no plan for doing it. I would like to move onto something other
than writing compilersOn Thursday, August 7, 2014 4:16:48 PM UTC-4, Yannick Duchêne wrote:
Le jeudi 7 août 2014 21:37:39 UTC+2, Martin DeMello a écrit :
A set of problems I just ran into:
- the repo at GitHub - githwxi/ATS-Postiats: ATS2: Unleashing the Potentials of Types and Templates is missing an
autogen.sh- when I ran through the steps manually, ./configure was generated but
running it did not generate a makefile- Makefile_devl seems to depend on an existing ats install:
Professor Xi may tell more, I just can say I remember he said: so far,
generating ATS2 C sources requires ATS1. Anyway, you will need an ATS
compiler installed to compile to C then compile C to binary. The original
sources are not C sources, they are ATS1 sources (and will be ATS2 sources
in the future).–
You received this message because you are subscribed to the Google Groups
“ats-lang-users” group.
To unsubscribe from this group and stop receiving emails from it, send an
email to ats-lang-user...@googlegroups.com.
To post to this group, send email to ats-lan...@googlegroups.com.
Visit this group at http://groups.google.com/group/ats-lang-users.
To view this discussion on the web visit
https://groups.google.com/d/msgid/ats-lang-users/9266984c-5dd5-40de-80c0-079049271092%40googlegroups.com.
one issue with open source type stuff is that unless there’s a
benevolent dictator to assign tasks, things get done in an order that
is random, subjective. that’s OK in some sense, don’t get me wrong.
eventually it hopefully converges on something more useful.
but…
but looking at it as a first time user, i think the ordering, the
priority, the biggest roi bang for the buck really needs to be what i
said before: (1) get rid of there being more than 1 canonical place to
go! (2) fix that to not have more than 1 way to build it! e.g. i have
no clue why the targz has to be different than what i get when cloning
at least for building. why is there no ./confgure in one but there is
in the other? just nutty incomprehensible ux to me.
so if anybody has the power to do (1) i think that would be really big
bang for the buck. obviously i can’t go and delete swaths of the web
site(s)… :-} and by ‘delete’ i mean coalesce so that there’s one
single correct easy to maintain place to keep, i do not mean to lose
valuable information of course.
$0.01 and dropping fast probably.
The original source of INSTALL is in the git-repo. Actually, all the files
in
sourceforge/ats-lang2 are copied from the git-repo. Anyway, it is not a big
deal. I can always merge changes into the git-repo if needed.On Thursday, August 7, 2014 12:47:18 AM UTC-4, Raoul Duke wrote:
(c) seems like the git repo version has other requirements: make
(yeah), automake (oh).
First and foremost, I really appreciate your long messages 
I do want to know other people’s perspectives.
one last (last? hah) suggestion: make a vm image and release it so
that i can just boot it up in some standard free open x86
virtualization environment (virtual box, or even
aws/openstack/whatever). that would be one nice kind of
end-run/gordian-knot solution.
We have already started a web-based system for programming with ATS.
But this will take time.
I certainly hope that more and more people will adopt ATS. However, I also
want to use ATS to build my own systems. I am willing to spend time pushing
for ATS but I cannot do it alone.
I understand how important smooth installation is meant for the adoption of
a language.
You could imagine that I received many criticisms in my course evaluation
precisely
because students had difficulty installing ATS (I should really thank the
person who wrote the
homebrew script as installing ATS on Mac had really been a black art)
In short, I hope more people would contribute. I also want to be seen as a
user of
ATS. The very point for me to develop ATS is that I want to use it. I do
not really want
to spend all of my time pushing for ATS.On Thursday, August 7, 2014 2:06:57 AM UTC-4, Raoul Duke wrote:
i can believe it – i’ve had plenty of technical debt situations to
face over the years. i’ve seen it both from the inside, where we are
‘comfortable’ / ‘inured’ / ‘beaten down’ / ‘too busy with other
things’ / ‘it works on my machine’ / etc. to deal with it. and i’ve
seen it from the outside, where i’m like, “ok never mind guess i’ll go
use something else now and never come back” and “oh the funnel never
got enough people and so now the product / company / nsf-sponsored
project is fired / dead / gone / will never get another 5 year renewal
grant, ever, period.”you are ‘comfortable’ with how it works.
new people are most likely less so. if at all. maybe actively not. so
this particular technical debt might be friction that slows adoption
of ats by anybody other than insiders. it is probably not very
quantifiable for many reasons, including that the people who give up
probably don’t have any interest in expending more energy telling you
about this; they just want to get on with life elsewhere.so anyway i think it is in your rational personal best interest to
find some way to address it sooner rather than later. i do not know if
you have worked in industry / with industry people, or have suffered
through Business Speak situations, but what you have here is A Problem
With Your Funnel. that’s baaaad mojo to people who actually care about
the bottom line, which in this day and age is usually adoption and
eyeballs (the theory being that eventually the money will follow :-).find an intern / summer student / undergrad who wants credit / out of
work hacker who needs to have something on their resume instead of
unemployed blankness. having done summer student / extra credit work
for PIs as an undergrad, and as having been a laid off software
engineer before, i expect that a university setting gives you more
access to such things… but maybe not.yes it is easy for an outsider such as meeeee to do a fly-by dropping
of advice. but this is not only me being an uptight easily annoyed
person who has really high standards for ui/ux (and who has already
used up any good will on cutting-edge languages long ago with bad haxe
or lfe-erlang installs, to name a few), it is also me being sad that i
see what i think are real basic up-front issues that are i dare say
slowing down ats adoption, which would be a crying shame.you know that in biology/biotech in academia there’s a push for the
software to be available and to work and for the data sets to be
available and to work and to give the same results? so while on the
one hand if you live in academia (i have, several times, on and off,
in my life) then you can sorta figure eh it doesn’t have to work for
anybody else. but i like that the bio community is saying, “back the
truck up, hold on a second! if the software doesn’t work for anybody
else, why should we believe the claims in the published paper?” and i
was sort of under the impression that you actually want ats to be more
than just something your direct reports use?i’m an outlier of course probably in terms of how i measure things up.
one last (last? hah) suggestion: make a vm image and release it so
that i can just boot it up in some standard free open x86
virtualization environment (virtual box, or even
aws/openstack/whatever). that would be one nice kind of
end-run/gordian-knot solution.(i suspect my $0.02 is probably not going to have a good exchange rate
with anybody around here any more
Thanks for the suggestions.
Eventually? Maybe.
For now, I hesitate. This requires a lot of work as many scripts need
to be modified for supporting the transition.On Thursday, August 7, 2014 1:21:15 AM UTC-4, Raoul Duke wrote:
ah.
ok so that leads me to suggest please: get rid of everything anywhere
other than github.(there’s a basic principle of usability here: make things as simple as
possible. entities must not be multiplied beyond necessity!)
- do not have sourceforge at all. this is like some kind of dmz
machine honeypot for making people waste their time? i actually hate
sourceforge and the crappy ui, and would much rather be doing pull
request ui on github in the first place. augh.- get rid of the install instructions on the ats web site. the might
be more detailed and even more correct, but ok then, put those updates
into the INSTALL file and the readme.md file on github so that there
is ONE single canonical place to get (and fix and correct and update)
the instructions.- try to not have any difference between the targz version and what i
would get from cloning. otherwise you waste people’s time trying to
figure out what is what where why when.- the targz can be a file to download from the github repo if need be.
- also that way, having the github readme.md, people who make e.g.
.deb files can update the readme with a pull request, to have an http
link directly to the download / instructions for things e.g. brew. i
mean i tried to use lauchpad but i have no freaking clue how to get
the deb files for ats2, where to find them, how to download them,
anything like that, because launchpad is just a horrible horrible
ui/ux.more $0.02.
© seems like the git repo version has other requirements: make
(yeah), automake (oh).
clarification: the pages mention it but
(a) they should mention it before the download steps
(b) the INSTALL file should mention it
© the pages say it is optional but doesn’t seem that way now. :-}
Correct in general.
and will be ATS2 sources in the future
I have no plan for doing it. I would like to move onto something other
than writing compilers :)On Thursday, August 7, 2014 4:16:48 PM UTC-4, Yannick Duchêne wrote:
Le jeudi 7 août 2014 21:37:39 UTC+2, Martin DeMello a écrit :
A set of problems I just ran into:
- the repo at GitHub - githwxi/ATS-Postiats: ATS2: Unleashing the Potentials of Types and Templates is missing an
autogen.sh- when I ran through the steps manually, ./configure was generated but
running it did not generate a makefile- Makefile_devl seems to depend on an existing ats install:
Professor Xi may tell more, I just can say I remember he said: so far,
generating ATS2 C sources requires ATS1. Anyway, you will need an ATS
compiler installed to compile to C then compile C to binary. The original
sources are not C sources, they are ATS1 sources (and will be ATS2 sources
in the future).
please do not let the perfect be the enemy of the
less-confusing-than-it-is-today. :-}
