Hi, 

I totally agreed to use -x instead to exclude the folder, this will also 
eliminate the extra unnecessary folder. Thanks for the better fix. 


As for the extra folder case, it could be very tricky as there are many 
possibilities of the dev packaging an module,  it could make the packaging 
flexible,  but also lacking of standardization?


Just my thought


Thanks


Rgds,
CL




---- On Tue, 05 Nov 2019 17:52:42 +0800 stephane.motte...@utc.fr wrote ----



Le 05/11/2019 à 08:18, Stéphane Mottelet a écrit :


Great analysis !


Le 5 nov. 2019 à 07:16, Chin Luh Tan <chinluh....@bytecode-asia.com> a écrit :



Hi,



just to add on to the issue on the atomsInstall. I finally find out that the 
issue was due to the extra "__MACOSX" folder created during zip process under 
MAC.



"The __MACOSX folder is created when a Mac user creates and archive (also 
called a zip file) using the Mac. If the Mac user sends the zip file to another 
Mac user, the folder will not appear - this is a hidden folder. "

 

During the atomsInstall process, the module zip file will be downloaded in the 
SCI\contrib\module_name\module_name.zip (or the SCIHOME atoms) and extract it.



In this stage, for windows and linux, it will be only one directory in 
SCI\contrib\module_name, namely module_name,

The packaging rules do not impose that the package root dir has the name 
"module_name". For example latest scicv has root dir "scicv-0.4".


If the zip or tar.gz contains more than one directory (by accident, or else) 
then it can break atomsExtract, on every platform. The atomsInstall process 
should end when such an incoherence is detected.


In the meantime (before the proposed above fix), to avoid repackaging all OSX 
zip packages it would be simpler to use the -x flag of unzip, e.g. like this


unzip IPCV-4.1-MacOS-bin.zip -x __MACOSX/*



    elseif regexp(archive_in,"/\.zip$/","o") <> [] then

        if getos() == "Windows" then
            extract_cmd = """" + 
getshortpathname(pathconvert(SCI+"/tools/zip/unzip.exe",%F)) + """";
        else
            extract_cmd = "unzip";
        end

        extract_cmd = extract_cmd + " -q -o """ + archive_in + """ -d """ + 
pathconvert(dir_out,%F) +"""";
       
        if getos() == "Darwin"
            extract_cmd = extract_cmd + " -x __MACOSX/*"
        end

    end

    [rep,stat,err] = unix_g(extract_cmd);


and it will be renamed to version number, such as 1.0, and the module 
installation would be successful. 



In Mac, it will have extra __MACOSX folder (unless the module packager manually 
remove it) and it will cause the system to detect this folder as the newly 
created module, and rename this to the version number, which the actual folder 
module_name, will remained as it is and gives error during module loading.

 

While guideline could be given to the modules developer to remove the _MACOSX 
folder before releasing it, we could also add the following line in 
atomsExtract function to ensure the renaming function find the correct folder 
instead of __MACOSX folder. 



atomsExtract.sci



    // Get the list of directories after the extraction

    // =========================================================================



    dirs_after = atomsListDir(dir_out);



    dirs_after(dirs_after=='__MACOSX') = [];   // <-- extra line proposed to add

   

    // Get the name of the created directory

    // =========================================================================



    dir_created = [];



    for j=1:size(dirs_after,"*")

        if find(dirs_after(j) == dirs_before) == [] then

            dir_created = dirs_after(j);

            break;

        end

    end

 



This shall solve the issue of module installation problem under MACOS.



Thanks.



Regards,
CL








---- On Tue, 05 Nov 2019 01:06:37 +0800 <stephane.motte...@utc.fr> wrote ----




There is a big difference in the inherited environment when scilab is launched 
from the command line and by clicking the icon.
In the past I solved a problem with the LANG variable

https://codereview.scilab.org/#/c/20787/
https://codereview.scilab.org/#/c/21069/

the second one is still not merged in master but has been included in 19 august 
build of Scilab 6.0.2 for OSX (the one available from www.scilab.org), since it 
was really blocking for many users.

Just to say that the regexp problem may be due to a similar problem (locale, 
encoding, ...)

S.

Chin Luh Tan <chinluh....@bytecode-asia.com> a écrit :

ok, i did:

 

1. Del both previous scilab and the scilab from your link installed just now to 
make it clean

2. Totally remove the .Scilab folder under home folder by using sudo rm -R 
.Scilab

3. Re-download the link 

4. Double click and install (by dragging the scilab icon to application folder)

5. run scilab from the app and type atomsList

 

Same regexp error until i launch it from terminal, and then it is ok. 

 

noted on the atoms issue for the modules with gateway, will try to look into it.

 

p/s: will continue any extra testing tomorrow. 

 

rgds,
CL
 

 

---- On Tue, 05 Nov 2019 00:24:53 +0800 <stephane.motte...@utc.fr> wrote ----

 



 






_______________________________________________
users mailing list
users@lists.scilab.org
http://lists.scilab.org/mailman/listinfo/users


Obviously, between points 3 and 4 I moved the Scilab app from the archive in 
the Appplication folder.

stephane.motte...@utc.fr a écrit :

Hello,

I don't have the regexp error. Here is my complete workflow:

1 - Delete the eventual previous Sciab 6.0.2 I had in the /Applications/ folder 
2 - Delete my .Scilab folder
3 - extract the .dmg I downloaded from 
http://www.utc.fr/~mottelet/scilab/download/6.0.2/scilab-6.0.2-x86_64-2019-11-04.dmg
4 - double click the Scilab icon (with first launch usual but annoying 
details...)
5 - click the ATOMS gui icon or type atomsList at Scilab prompt

Concerning the problem with IPCV (and maybe other packages with gateways, I 
only tested IPCV), I did not address it in the above archive. I only modified 
the scilab script (I will create a commit on codereview.scilab.org) and removed 
the libcurl.dylib that was shipped in lib/thirdparty folder.

Best,

S.

Chin Luh Tan <chinluh....@bytecode-asia.com> a écrit :


Hi, 

 

from my end:

 

1. The warning for the 10.15 gone. 

2. First launch from the app still give the regexp error when running 
atomsList, but after running one time from terminal and get it run 
successfully, I could run it from the app as after all. 

3. atomsInstall for the module with gateway folder still extracted to 2 
separate folders which need to be manually move in order for it to run. No 
issue for macros only module. 

 

Thanks for the great effort. Is is possible to share the compilation steps? 

 

rgds,

CL

 

 

---- On Mon, 04 Nov 2019 22:40:20 +0800 <stephane.motte...@utc.fr> wrote ----

 


Did you test the Atoms stuff ?

Heinz Nabielek <heinznabie...@me.com> a écrit :

On 04.11.2019, at 14:55, stephane.motte...@utc.fr wrote:


Thanks Samuel. So the regexp error can eventually have the same cause (libcurl 
version mismatch). I have packaged a scilab-6.0.2 without these libraries (and 
removing the 10.15 warning) for OSX users at the following URL:

https://antispam.utc.fr/proxy/1/c3RlcGhhbmUubW90dGVsZXRAdXRjLmZy/www.utc.fr/~mottelet/scilab/download/6.0.2/scilab-6.0.2-x86_64-2019-11-04.dmg

CL, Heinz, Willi can you test it ?

Works to perfection and without this funny warning.

 

 


I recommend temporary renaming/moving your .Scilab/scilab-6.0.2 directory 
before testing


Thanks. Very helpful. Great suggestion. I would not have thought of it…..

 

Thanks for your dedication to this work.Heinz



 

_______________________________________________

users mailing list

users@lists.scilab.org

http://lists.scilab.org/mailman/listinfo/users

 



 


 

_______________________________________________

users mailing list

users@lists.scilab.org

http://lists.scilab.org/mailman/listinfo/users




_______________________________________________
users mailing list
users@lists.scilab.org
https://antispam.utc.fr/proxy/1/c3RlcGhhbmUubW90dGVsZXRAdXRjLmZy/lists.scilab.org/mailman/listinfo/users



_______________________________________________
users mailing list
users@lists.scilab.org
https://antispam.utc.fr/proxy/1/c3RlcGhhbmUubW90dGVsZXRAdXRjLmZy/lists.scilab.org/mailman/listinfo/users

-- 
Stéphane Mottelet
Ingénieur de recherche
EA 4297 Transformations Intégrées de la Matière Renouvelable
Département Génie des Procédés Industriels
Sorbonne Universités - Université de Technologie de Compiègne
CS 60319, 60203 Compiègne cedex
Tel : +33(0)344234688
http://www.utc.fr/~mottelet

_______________________________________________
users mailing list
users@lists.scilab.org
http://lists.scilab.org/mailman/listinfo/users
_______________________________________________
users mailing list
users@lists.scilab.org
http://lists.scilab.org/mailman/listinfo/users

Reply via email to