On 2013-06-07, at 20:03, [email protected] wrote:

> Marcus Denker wrote:
>> Hi,
>> 
>> I changed all links to point to the latest 2.0 build on files.pharo.org
>> 
>> Reason:
>> 
>>      -> We are fixing bugs, yet people downloading 2.0 do not see the fix
>>      -> We had multiple times that people wasted quite some effort 
>> re-reporting bugs
>>             that are already fixed.
>> 
>> Notes
>>      -> the file name "latest.zip" is not good for the image. It should 
>> encode that it is the image only and the version number maybe?
>>      -> I did not change any VM related download links, but I think we 
>> should do the same there.
>> 
>> 
>>      Marcus
>>  
> That is great that 2.0 is continuing to improve.
> 
> +100 for including version numbers on the web page - something like [1].  For 
> the purpose of reproducibility and least surprise, at [2] where it currently 
> says "2.0 (release)"  and "Pharo2.0 (win)" I would expect that link to 
> download the _exact_ same file every time, except today that zip file 
> contains files dated today, so obviously is not the same file as last week.  
> You might get away with following "latest" with a link called "Pharo2 (win)" 
> without the minor-version number, but by including a a minor-version number I 
> would expect that it is the constant point-in-time snapshot of the _2.0 
> release_. 
> [1] http://www.python.org/download/
> [2] http://www.pharo-project.org/pharo-download
> 
> Also, revisiting (major.minor.bugfix) semantic versioning in the context of 
> Pharo, an important part of configuration management for a user is matching 
> versions of VM to Image.  For me, I simply this when I update by always 
> taking the fat zip which include both - but it would be nice to have the 
> confidence to just take the image.  So it would be good if the VM/Image 
> dependency could be encoded by the version numbers.  Something like the 
> following...
> 
> My shallow understanding of the CI infrastructure is that to "some degree" :
> 1. the Image CI runs against a constant VM version

the image is tested against the latest stable vm (ex. 
http://files.pharo.org/vm/pahro/linux/stable.zip)
since stable VMs are promoted manually we can consider them being constant.

> 2. the VM CI runs against a constant Image version

true, https://ci.inria.fr/pharo/view/VM/job/PharoVM-tests/ shows that we run 
against the latest image (2.0 in this case) and a fixed image (vm-test.zip in 
this case). Currently the CI infrastructure is suffering from a network 
failure, that's why the more runs than usual are red.


> So considering Case 1...
> a. Image updates would bump only the bugfix number. b. VM updates would bump 
> the minor number.
> This would make is easy for a user to always determine that Image version 
> 2.2.x was tested against VM Pharo22.exe. 
> To avoid excessively incrementing the minor version number, considering Case 
> 2...
> c. (for example) you would begin testing VM updates against constant Image 
> 2.0.0
> d. when a required VM update breaks a test in Image 2.0.0, then increment the 
> minor version number.  Work to get latest VM and latest Image both green.  
> These then become Pharo21.exe and Image 2.1.0, which are used as a the next 
> starting point for Case 1.  Further VM updates are now tested against 
> constant Image 2.1.0.

this scheme will work fine once we release VM's for each Pharo release (which I 
guess is planned at some point).
Currently this is not yet the case, so the objective is to make the VM work 
with every image...


Reply via email to