Re: [racket-dev] package system, minimal builds and more

2013-10-09 Thread Neil Toronto 

On 10/01/2013 07:30 AM, Neil Toronto wrote:

On 10/01/2013 09:20 AM, Tobias Hammer wrote:

* monolithic math
currently math is one big package and installing it pulls in nearly
everything through the docs. Is it planned to split it into -lib and
-doc?


We were waiting for a reason. This is one. I'll try it on my flight home
tonight. (Never done any package management before.)


Okay, that took longer than I thought. It was much easier than splitting 
the "plot" package, though...


Neil ⊥

_
 Racket Developers list:
 http://lists.racket-lang.org/dev

Re: [racket-dev] package system, minimal builds and more

2013-10-07 Thread Tobias Hammer 

All in one answer

On Tue, 01 Oct 2013 15:30:11 +0200, Neil Toronto   
wrote:

On 10/01/2013 09:20 AM, Tobias Hammer wrote:

* monolithic math
currently math is one big package and installing it pulls in nearly
everything through the docs. Is it planned to split it into -lib and  
-doc?


We were waiting for a reason. This is one. I'll try it on my flight home  
tonight. (Never done any package management before.)


Thanks for tackling this. I think its worth the effort.

I suppose this is a good rule of thumb: if your docs make pictures, put  
them in a separate package.


Yes, thats a good minimum requirement. For standalone builds or for small  
devices it
would be even better if it's possible to build everything that might be  
useful in

this setting without a single doc.


On Thu, 03 Oct 2013 17:26:08 +0200, Robby Findler  
 wrote:



These two are fixed now (with the snapshot build
http://plt.eecs.northwestern.edu/snapshots/20131003-154d940/).


Thanks, will try as soon as possible.


On Wed, 02 Oct 2013 16:39:42 +0200, Sam Tobin-Hochstadt  
 wrote:



* raco pkg warnings
On every command i get
warning: tool "pkg" registered twice: "pkg" and
#
warning: tool "setup" registered twice: "setup" and
#
warning: tool "link" registered twice: "link" and
#
 /home/hamm_to/tmp/racket/_tmp/racket-5.90.0.9/ is the used installation.


What command did you run to get this error? I haven't been able to  
replicate it.


It seems to only happen if i have "PLTCOLLECTS=:", not if it's set to  
anything else to unset.

Then a "raco pkg" is enough to trigger it.



You can see these with `raco pkg show`.  Is that what you're looking for?


Exactly what i was looking for (raco pkg show -a -d). Thanks.


On Wed, 02 Oct 2013 18:32:51 +0200, Jay McCarthy   
wrote:



I think I need more details.

If you have pkg A installed, then it should only download a small
checksum to see if A needs to be updated. If this is broken, then it
is an error.


That works as supposed.


If you do NOT have pkg A installed, then it will download the full
file. If that install fails, then 'raco pkg' cleans up after itself
and deletes the things it downloaded. Apt does not do this and saves
everything in a temporary directory that must be manually cleaned. I
did not implement that because it feels wrong to run a command like
'raco pkg clean-old-tmp-files' or something. But it sounds like you
want that?


Yes, that's what i had in mind because I'm used to this behavior and i was
wondering why raco downloaded the files over and over on a _failed_ install
(Only noticed it because it was a relatively large install). But the  
temporary
files are a good point that i completely forgot. I think, in hope that  
installs

(hopefully) rarely fail it's better as it is now.

Tobias
_
 Racket Developers list:
 http://lists.racket-lang.org/dev

Re: [racket-dev] package system, minimal builds and more

2013-10-03 Thread Robby Findler 
On Tue, Oct 1, 2013 at 8:20 AM, Tobias Hammer  wrote:

> Hi,
>
> i played a bit with the new package system, minimal installs and cross
> compiling. It works pretty good but i have a few problems and
> remarks/questions.
> I start with a fresh yesterdays minimal source
> (min-racket-5.90.0.9-src.tgz), build it (natively for linux) and try to
> install packages.
>
> (Given in no particular order)
>
>
> * compiler-lib install fails with the following error
> ./bin/raco pkg install -i compiler-lib
> Resolving "compiler-lib" via http://www.eecs.northwestern.**
> edu/plt/snapshots/20130929-**6d4ff30/catalog/
> Resolving "compiler-lib" via https://pkg.racket-lang.org
> Downloading checksum for compiler-lib
> Downloading http://racket-packages.s3-us-**west-2.amazonaws.com/pkgs/**
> b824ed762533834659b5f0ee047dea**fb101acc74/compiler-lib.zip
> raco pkg install: package conflicts with existing installed
>   package: compiler-lib
>   module path: compiler/option
>
> As its needed by racket-doc and similar it blocks install of a lot of
> packets
>
>
> * the snapshot catalog seems to be nonexistent
> If i browse to http://www.eecs.northwestern.**edu/plt/snapshots/20130929-*
> *6d4ff30/catalog/,
> i always get a 404. Is this intended?
>
>

These two are fixed now (with the snapshot build
http://plt.eecs.northwestern.edu/snapshots/20131003-154d940/).

Thanks!

Robby
_
  Racket Developers list:
  http://lists.racket-lang.org/dev

Re: [racket-dev] package system, minimal builds and more

2013-10-02 Thread Jay McCarthy 
On Wed, Oct 2, 2013 at 8:39 AM, Sam Tobin-Hochstadt
 wrote:
>> * redownload after fail
>> the package manager seems to download every packet again after a failed
>> install or user interruption. Would it be worth to reuse the once downloaded
>> zips if the checksum is the same (similar to the behavior of apt).
>
> I believe the package manager knows how to do this, but it doesn't
> seem to be doing it here, even though there's checksum information in
> the catalog files. Jay, do you know what's going on here?

I think I need more details.

If you have pkg A installed, then it should only download a small
checksum to see if A needs to be updated. If this is broken, then it
is an error.

If you do NOT have pkg A installed, then it will download the full
file. If that install fails, then 'raco pkg' cleans up after itself
and deletes the things it downloaded. Apt does not do this and saves
everything in a temporary directory that must be manually cleaned. I
did not implement that because it feels wrong to run a command like
'raco pkg clean-old-tmp-files' or something. But it sounds like you
want that?

Jay

-- 
Jay McCarthy 
Assistant Professor / Brigham Young University
http://faculty.cs.byu.edu/~jay

"The glory of God is Intelligence" - D&C 93
_
  Racket Developers list:
  http://lists.racket-lang.org/dev

Re: [racket-dev] package system, minimal builds and more

2013-10-02 Thread Sam Tobin-Hochstadt 
On Tue, Oct 1, 2013 at 9:20 AM, Tobias Hammer  wrote:
> Hi,
>
> i played a bit with the new package system, minimal installs and cross
> compiling. It works pretty good but i have a few problems and
> remarks/questions.
> I start with a fresh yesterdays minimal source
> (min-racket-5.90.0.9-src.tgz), build it (natively for linux) and try to
> install packages.
>
> (Given in no particular order)

First, thanks for the feedback!

> * compiler-lib install fails with the following error
> ./bin/raco pkg install -i compiler-lib
> Resolving "compiler-lib" via
> http://www.eecs.northwestern.edu/plt/snapshots/20130929-6d4ff30/catalog/
> Resolving "compiler-lib" via https://pkg.racket-lang.org
> Downloading checksum for compiler-lib
> Downloading
> http://racket-packages.s3-us-west-2.amazonaws.com/pkgs/b824ed762533834659b5f0ee047deafb101acc74/compiler-lib.zip
> raco pkg install: package conflicts with existing installed
>   package: compiler-lib
>   module path: compiler/option
>
> As its needed by racket-doc and similar it blocks install of a lot of
> packets

The problem here seems to be that the below problem causes it to fall
back to an older version from the central pkg catalog, which doesn't
work because of refactoring between when your snapshot was from, and
when the upload of that zip file to the pkg catalog happened.

Once we actually release the next version, we won't move things around
in ways that cause these compatibility issues.

I manually edited etc/config.rktd to point to the Utah snapshot page,
and then it worked.

> * the snapshot catalog seems to be nonexistent
> If i browse to
> http://www.eecs.northwestern.edu/plt/snapshots/20130929-6d4ff30/catalog/, i
> always get a 404. Is this intended?

I think this is a bug in how the Northwestern snapshots are set up,
because they should go to 'plt.eecs.northwestern.edu/snapshots/...'
instead.  I've cc'ed Robby, who can hopefully fix this.

> * redownload after fail
> the package manager seems to download every packet again after a failed
> install or user interruption. Would it be worth to reuse the once downloaded
> zips if the checksum is the same (similar to the behavior of apt).

I believe the package manager knows how to do this, but it doesn't
seem to be doing it here, even though there's checksum information in
the catalog files. Jay, do you know what's going on here?

> * raco pkg warnings
> On every command i get
> warning: tool "pkg" registered twice: "pkg" and
> #
> warning: tool "setup" registered twice: "setup" and
> #
> warning: tool "link" registered twice: "link" and
> #
>
> /home/hamm_to/tmp/racket/_tmp/racket-5.90.0.9/ is the used installation.

What command did you run to get this error? I haven't been able to replicate it.

> * More verbose require error
> If i require something nonexistent i get an exception like
> (require fdfsdfsdf)
> ; readline-input:4:8: collection not found
> ;   for module path: fdfsdfsdf
> ;   collection: "fdfsdfsdf"
> ;   in collection directories:
> ;/home/hamm_to/.racket/snapshot/collects
> ;/volume/USERSTORE/hamm_to/tmp/racket-qnx/linux-racket-5.90.0.9/collects
> ;/home/hamm_to/.racket/snapshot/collects
> ;/volume/USERSTORE/hamm_to/tmp/racket-qnx/linux-racket-5.90.0.9/collects
> ;... [22 additional linked and package directories]
>
> Is it possible to somehow show these additional 22 entries? Maybe not in the
> error but from raco?

You can see these with `raco pkg show`.  Is that what you're looking for?

Sam
_
  Racket Developers list:
  http://lists.racket-lang.org/dev

Re: [racket-dev] package system, minimal builds and more

2013-10-01 Thread Neil Toronto 

On 10/01/2013 09:20 AM, Tobias Hammer wrote:

* monolithic math
currently math is one big package and installing it pulls in nearly
everything through the docs. Is it planned to split it into -lib and -doc?


We were waiting for a reason. This is one. I'll try it on my flight home 
tonight. (Never done any package management before.)


I suppose this is a good rule of thumb: if your docs make pictures, put 
them in a separate package.


Neil ⊥

_
 Racket Developers list:
 http://lists.racket-lang.org/dev

[racket-dev] package system, minimal builds and more

2013-10-01 Thread Tobias Hammer 

Hi,

i played a bit with the new package system, minimal installs and cross  
compiling. It works pretty good but i have a few problems and  
remarks/questions.
I start with a fresh yesterdays minimal source  
(min-racket-5.90.0.9-src.tgz), build it (natively for linux) and try to  
install packages.


(Given in no particular order)


* compiler-lib install fails with the following error
./bin/raco pkg install -i compiler-lib
Resolving "compiler-lib" via  
http://www.eecs.northwestern.edu/plt/snapshots/20130929-6d4ff30/catalog/

Resolving "compiler-lib" via https://pkg.racket-lang.org
Downloading checksum for compiler-lib
Downloading  
http://racket-packages.s3-us-west-2.amazonaws.com/pkgs/b824ed762533834659b5f0ee047deafb101acc74/compiler-lib.zip

raco pkg install: package conflicts with existing installed
  package: compiler-lib
  module path: compiler/option

As its needed by racket-doc and similar it blocks install of a lot of  
packets



* the snapshot catalog seems to be nonexistent
If i browse to  
http://www.eecs.northwestern.edu/plt/snapshots/20130929-6d4ff30/catalog/,  
i always get a 404. Is this intended?



* redownload after fail
the package manager seems to download every packet again after a failed  
install or user interruption. Would it be worth to reuse the once  
downloaded zips if the checksum is the same (similar to the behavior of  
apt).



* raco pkg warnings
On every command i get
warning: tool "pkg" registered twice: "pkg" and  
#
warning: tool "setup" registered twice: "setup" and  
#
warning: tool "link" registered twice: "link" and  
#


/home/hamm_to/tmp/racket/_tmp/racket-5.90.0.9/ is the used installation.


* sometimes installations fail without a reason or hangs
./bin/raco pkg install -i xrepl
[...]
The following uninstalled packages are listed as dependencies of  
macro-debugger-text-lib

and they will be installed:
   parser-tools-lib
   unstable-list-lib
Resolving "parser-tools-lib" via  
http://www.eecs.northwestern.edu/plt/snapshots/20130929-6d4ff30/catalog/

Resolving "parser-tools-lib" via https://pkg.racket-lang.org
Resolving "parser-tools-lib" via https://planet-compat.racket-lang.org
raco pkg install: cannot find package on catalogs
  package: parser-tools-lib

rerunning the command and it suddenly works.


* monolithic math
currently math is one big package and installing it pulls in nearly  
everything through the docs. Is it planned to split it into -lib and -doc?



* cross compiling and installing
is it possible to install packages as part of the bootstrap process? the  
problem i have is that the build machine is a regular linux but the cross  
compile target is constrained in that it cannot install packages (no  
internet, no writable discs, ...). But i can't install via raco because  
racket is compiled for that target it does not run on linux.
I can currently work around this by installing a linux racket with same  
version, installing there and then copying share/, but thats somewhat ugly.



* More verbose require error
If i require something nonexistent i get an exception like
(require fdfsdfsdf)
; readline-input:4:8: collection not found
;   for module path: fdfsdfsdf
;   collection: "fdfsdfsdf"
;   in collection directories:
;/home/hamm_to/.racket/snapshot/collects
; 
/volume/USERSTORE/hamm_to/tmp/racket-qnx/linux-racket-5.90.0.9/collects

;/home/hamm_to/.racket/snapshot/collects
; 
/volume/USERSTORE/hamm_to/tmp/racket-qnx/linux-racket-5.90.0.9/collects

;... [22 additional linked and package directories]

Is it possible to somehow show these additional 22 entries? Maybe not in  
the error but from raco?



I know these are a lot of things and some might be obvious or result from  
misunderstanding or missing something. So thanks for enlightening answers  
to any of these.


Tobias
_
 Racket Developers list:
 http://lists.racket-lang.org/dev