>IMO, the binary package can only contain pbj's if the source package contains 
>the source pbk's and a way to compile them. 
>The binary package probably cannot contain >a zip of the pbj's either.  The 
>idea of the binary package is that it contains the result of the compilation, 
>to save folks >having to set up the compile/build environment.  >My 
>understanding is that it cannot contain other artifacts.

You are correct.
So does this mean the pbk would be duplicated on both dist repo and flex-sdk 
git repo?  
I don't like the idea to have duplicate source code.

Otherwise, this means the pbk would be in one single place, so necessarily in 
flex-sdk git repo  (as is currently the case).
This implies the build file in dist would fetch the pbk sources from git, 
which means cloning the whole git flex-sdk repo for 10 source files...

So it's git clone for 10 files, or duplicate source, but for files that won't 
change much.

WDYT?

Maurice 

-----Message d'origine-----
De : Alex Harui [mailto:aha...@adobe.com] 
Envoyé : jeudi 12 décembre 2013 18:20
À : dev@flex.apache.org
Objet : Re: PixelBender and Builds.a.o



On 12/12/13 9:00 AM, "Maurice Amsellem" <maurice.amsel...@systar.com>
wrote:

>> B) I don't think I understood your 4a, but it doesn't sound like it 
>>would meet policy.  No zips are allowed in svn.
>
>I am confused. 
> Isn't
>https://dist.apache.org/repos/dist/dev/flex/sdk/4.11.0/rc1/binaries/  
>an svn repo?
> And it does contain zip and binaries.
>
>So if svn cannot contain zips, how do we make commit the result of pb 
>compilation ? as plain files ?
Ah, I understand now.  Yes, we use SVN to put things on dist.  We didn't used 
to.  When you say svn or git I think of our repos.  Dist (and
archive) can have binary artifacts.
>
>My 4a) is more or less your B) =>  "build scripts copy the pbk source 
>from dist into the source package and pbjs from dist into the binary 
>package".
>Expect that I retrieve one zip file with internal directory structure 
>(same as the one produced by obsolete pixel_bender Jenkins job) from 
>dist and copy it to binary package, instead of individual flat files.
>I don't think this breaks any rule ..
IMO, the binary package can only contain pbj's if the source package contains 
the source pbk's and a way to compile them.  The binary package probably cannot 
contain a zip of the pbj's either.  The idea of the binary package is that it 
contains the result of the compilation, to save folks having to set up the 
compile/build environment.  My understanding is that it cannot contain other 
artifacts.

>
>>If you plan to take this on, please let us know
>Not at the moment.  I am just trying to detail the steps, in case it 
>needs to be done later (by anyone who want to).
>Hoping the infra will fix the current issue...
>
>Maurice
>
>-----Message d'origine-----
>De : Alex Harui [mailto:aha...@adobe.com] Envoyé : jeudi 12 décembre 
>2013 17:44 À : dev@flex.apache.org Objet : Re: PixelBender and 
>Builds.a.o
>
>I agree with most of it except:
>
>A) we can't auto-commit to dist.  I believe only voted on artifacts can 
>go there.  But realistically, I think we're only going to commit once 
>unless someone find a reason to change the PBK files.
>
>B) I don't think I understood your 4a, but it doesn't sound like it 
>would meet policy.  No zips are allowed in svn.
>
>A binary package is only supposed to additionally contain compiled 
>artifacts of the source package. I'd rather not have to change the 
>installer either right now since I'm working on a major upgrade to it 
>and don't want to add that to the delays to get our CI up and running again.
>I think the best short term answer is that the SDK build scripts copy 
>the pbk source from dist into the source package and pbjs from dist 
>into the binary package.  That sort of fudges the tag==package policy 
>but essentially we can show there are already multiple tags in our 
>packages since we pull TLF from its own repo.
>
>If you plan to take this on, please let us know.  I'm waiting to see if 
>Infra did try again to reboot the server. If they did and the recent 
>failure means it is still busted, it is definitely time to start making 
>these changes, so one of should start on it.
>
>-Alex
>
>On 12/12/13 8:27 AM, "Maurice Amsellem" <maurice.amsel...@systar.com>
>wrote:
>
>>Thanks Alex.
>>
>>So IIUC,
>>
>>1) create new folder "pixel_bender" under 
>>https://dist.apache.org/repos/dist/dev/flex/  SVN repo
>>    - will contain the zip files generated by the build below
>>
>>2) add new project " flex-sdk_pixelbender" in "flex-utilities" GIT  repo
>>containing:   
>>    - PBK sources for flex-sdk
>>    - build file for compiling pbk into pbj
>>      Output = pb.zip file containing pbj + pbk (similar structure to 
>>current pixelbender upstream Jenkins job)
>>             [Bonus] auto-commmit  the zip file to 
>>dist/.../pixel_bender
>>
>>=> build file must be run manually by "Release manager"
>>
>>3) remove flex-sdk_pixelbender upstream job from b.a.o Jenkins
>>(obsolete)
>>
>>4) to include  pbk+pbj in Flex SDK  release there are two options
>>a) modify build_release.sh (or equivalent) to checkout pb.zip from svn 
>>and unzip into dist
>>b) add checkout+checkMD5+unzip step in flex sdk installer
>>
>>Is that correct?
>>What option in 4)  I think a) is the easiest, and it does not break 
>>any Apache rule.
>>
>>Maurice
>>
>>-----Message d'origine-----
>>De : Alex Harui [mailto:aha...@adobe.com] Envoyé : jeudi 12 décembre
>>2013 15:40 À : dev@flex.apache.org Objet : Re: PixelBender and 
>>Builds.a.o
>>
>>
>>
>>On 12/12/13 5:03 AM, "Maurice Amsellem" <maurice.amsel...@systar.com>
>>wrote:
>>
>>>I am trying to understand the steps below:
>>>
>>>Where would the pbk => pbj compilation take place ?
>>In the build script for this "project".
>>> 
>>>Where would the pbj be stored after the build , if the build does not 
>>>occur on the CI?
>>On the same servers we store our voted on releases.
>>>Will the pbj be compiled by Jenkins ? or built manually and stored 
>>>somewhere ?
>>All releases are compiled by someone running the build script and 
>>signing the artifacts.  Apache even seems to not want to allow signing 
>>of Jenkins-built artifacts.
>>
>>-Alex
>>
>

Reply via email to