'If our tests require Flash, then there is no point in running them if
there are no SWF artifacts.'
This is just an info dump, in case the following is useful:
(short version) My failing combination is a) FLASHPLAYER_DEBUGGER is
defined and b) PLAYERGLOBAL_HOME is not defined and/or local m2 does not
contain playerglobal swc
But this is just with standard maven build, not with profiles or anything
specified. Also on windows, in case that matters in any way.
(Details)
I am still getting to grips with 'profiles' in maven. I probably read about
that at one point and have used them, but will go back to refresh my
knowledge (now that I understand more of the basics).
I did observe that the MavenTestAdapter has a getPlayerGlobal method which
looks for the swc in the tests for compiler (not compiler-jx), this still
tries to run the tests if playerglobal is missing (and assuming the debug
player is available), but the player has bad bytecode (e.g. error dialogs
like TypeError: Error #2023: Class ASDateTests1933741105634631672$ must
inherit from Sprite to link to the root.)
When I look at getPlayerglobal() inside MavenTestAdapter I do see something
a little confusing (to me).
It checks to see if PLAYERGLOBAL_HOME is defined. If it is not, it bails
and returns null. This will cause an error later in testing
But PLAYERGLOBAL_HOME does not appear to be used to find the playerglobal
swc anywhere else, so this check may not even be relevant.
if PLAYERGLOBAL_HOME is defined, then it ignores it and looks for the swc
in System.getProperty("mavenLocalRepoDir"). This is why I think it finds it
after it has been cached in local m2 and perhaps why things continue to
work when it is removed from the pom dependencies.
However, if I unset FLASHPLAYER_DEBUGGER env var then the tests phase
passes and the build continues on to completion.
So my failing combination is a) FLASHPLAYER_DEBUGGER is defined and b)
PLAYERGLOBAL_HOME is not defined and/or local m2 does not contain
playerglobal swc
On Fri, May 24, 2019 at 7:33 AM Alex Harui <[email protected]> wrote:
> I guess I am not making a clear statement. I understand you are trying to
> help others, but unless you have tested from scratch with both generating
> SWF artifacts and not generating SWF artifacts then you haven't actually
> helped everyone, just those who want the same set of artifacts you are
> expecting.
>
> I would not expect any solution to include an Adobe artifact without using
> a profile to include it.
>
> You might need two videos, one for generating SWF artifacts and one for
> not.
>
> If our tests require Flash, then there is no point in running them if
> there are no SWF artifacts.
>
> Thanks,
> -Alex
>
> On 5/23/19, 11:06 AM, "Carlos Rovira" <[email protected]> wrote:
>
> Hi Alex,
>
> I'm not getting this working only for me, in fact my motivation was
> exactly
> the opposite. The final motivation is to able to do a short video to
> post
> on an Apache Royale youtube channel, since is something many, many
> people
> requested. And something I think will give us more users and exposure.
>
> I was working without problem each day. I tried to remove repository
> folder
> to simulate a "day 0" like a new user to see if all was working. The
> result
> was it was failing.
>
> Now with a dependency added in compiler's pom and a profile added in
> the
> wiki instructions, I was able to build from scratch. Other's can try
> this
> to proof is a solution for anyone.
>
> IOW, If a new user tries the wiki steps some days ago he'd found royale
> didn't build, and fails with the error exposed here, and will get
> stuck.
> Now, hopefully he will get it working.
>
> For me is ok, all you say (maybe the only thing I don't agree is put
> skipTests to false as an official way to make maven build work
> officially,
> since in maven tests are mandatory, and you must opt-out, with a
> profile o
> via command line, but official build should work with normal tests in a
> first execution).
>
> About having a repository or not: This should not matter, but the fact
> is
> it currently does, independently of what any of us want. I, as you,
> would
> want the simplest way to build, that could be always the same, but
> there's
> a difference in a first maven build of royale against the subsequent
> builds, that can be simplified (removing the -s settings... and the
> -Pprofile..). I didn't design the process, but is what we have now. So
> is
> important to test against an empty repository folder, unless we change
> the
> build process and get it more simpler, what I don't expect to happen
> anytime soon, since all of us have many things on plate right now.
>
> I think we all understand the goals, and that we have two sets of
> outputs.
> Right now, I only know how to get one of them. If there's other one you
> know you can post it here and I can put on wiki, or directly modify
> wiki
> page that is the official one. If you can do the second, it would be
> great
> since it will be more accurate to what you have in mind.
>
> If you have no time fo now what we can do is:
> 2
> a) I can reintroduce the "-Putils" line in the wiki as something to do
> in a
> concrete case, since right now (at the time we are writing this), as
> you
> posted is important in a concrete situation, but not in building from
> scratch (for now until your changes will be merged).
>
> b) As soon as you get your branch working and merged in develop, you
> should
> change the wiki to conform to the needs of the changes you will
> introduce
> in your branch. I'll be interested in give a hand here and test it
> againts
> an empty repo, and from a Mac, and help to refine the process and the
> wiki
> if needed.
>
> About the planned video, since is a time consuming work maybe better to
> postpone until your work is merged so I can create one that doesn't get
> obsolete in few days.
>
> It's ok for you?
>
> thanks
>
>
>
> El jue., 23 may. 2019 a las 18:29, Alex Harui
> (<[email protected]>)
> escribió:
>
> >
> >
> > On 5/23/19, 3:04 AM, "Carlos Rovira" <[email protected]>
> wrote:
> >
> > Hi Alex,
> >
> > El jue., 23 may. 2019 a las 3:49, Alex Harui
> (<[email protected]
> > >)
> > escribió:
> >
> > > Before we go too far in any one direction, I may not be able to
> > respond
> > > fully to this thread today as there seems to be a lot to catch
> up
> > on, but
> > > let me try to summarize the goals of the Maven build.
> > >
> > > 1) There are some helper jars (compiler-build-tools and
> > > compiler-jburg-types). They are built by the "utils"
> profile. They
> > > haven't changed in develop, but they will change in 0.9.6.
> They've
> > been
> > > changed in the release_practice branch. So folks will need to
> use
> > the
> > > "utils" profile whenever we (rarely) change those jars.
> > >
> > >
> > Ok, so we should put in wiki that utils profile is needed for
> that
> > case,
> > but not for "initial" build case. I'm worried to try to simplify
> > instructions and process to minumun needs to avoid new comers
> > confusion.
> > So, I'll mention utils profile as a special case to execute when
> > needed.
> >
> > As soon as I merge release_practice into develop, you will need to
> use the
> > utils profile to build from scratch.
> >
> > > 2) Adobe will probably never publish official playerglobal on
> Maven.
> > > There is a whole bunch of logic in the Mavenizer to address
> licensing
> > > acceptance issues.
> > >
> >
> > For what we discussed in the thread, seems playerglobal is
> already on
> > maven
> > official repos, so my guest is we are served with that and don't
> need
> > adobe
> > host it in a maven repo.
> >
> > Adobe has not given permission to distribute playerglobal in this
> way so
> > we cannot use it.
> >
> > >
> > > 3) IIRC, the most recent changes were to allow the Maven build
> to
> > work
> > > without requiring SWF versions of artifacts and probably
> > > playerglobal/airglobal. So, adding hard requirements to
> > playerglobal will
> > > defeat this capability unless those dependencies are in the
> > appropriate
> > > Maven profile.
> > >
> >
> > Right now we need to do this:
> >
> > mvn -s settings-template.xml clean install
> -Pgenerate-swcs-for-swf
> > so this means something is not working ok in a clean environment
> for
> > first
> > build/install?
> > For now, the current instructions works, but if that's the case,
> we
> > should
> > try to fix this in the future, although seems this is not urgent
> while
> > people is capable of build Royale in the current way.
> >
> > The goal for Maven, like the goal for the Ant builds, is to not
> require
> > Adobe artifacts and build JS-only versions. Building SWF versions is
> > opt-in. I'm not surprised there are bugs after these changes, but
> the
> > solutions should consider that there are two different sets of
> output.
> >
> > >
> > > 4) The CI builds (builds.a.o and apachroyalecibuild) are good
> > reference
> > > examples of Maven building things correctly on Windows. You
> can
> > compare
> > > your setup and console output to those builds.
> > >
> >
> > I was building without problem and still can build without
> problem. My
> > concern was for the case people tries to build maven for the
> first
> > time,
> > and was where I found problems. This problems are as well not
> > reproduced in
> > machines that are already working, since they pass the initial
> setup.
> >
> >
> > >
> > > 5) There might be some assumption that airglobal and/or
> playerglobal
> > exist
> > > to determine whether the build is going to try to output SWF
> > versions of
> > > the artifacts or not.
> > >
> > > 6) The default, IIRC, is to not require airglobal/playerglobal
> and
> > build a
> > > JS-Only set of artifacts similar to how it is done in the Ant
> builds.
> > >
> >
> > So, this wiki walkthrough:
> >
> >
> https://nam04.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub.com%2Fapache%2Froyale-asjs%2Fwiki%2FBuild-Apache-Royale-with-Maven&data=02%7C01%7Caharui%40adobe.com%7Cb2aa222889e146d467ab08d6dfa95bd1%7Cfa7b1b5a7b34438794aed2c178decee1%7C0%7C0%7C636942315815634189&sdata=P0y3B4j%2Bf7Tg%2BtdB%2FjS%2BlbdBrTOrmUzgRnTFeOHMwEc%3D&reserved=0
> > is describing whole process without differentiation.
> > can be updated to build with maven SWF/JS and in the other hand
> only
> > JS?
> > I think the actual page description us for SWF/JS, and I
> personally
> > never
> > try / or know how to build just JS, what would be very
> interesting
> > since
> > many people will really only build for JS, and if sometime in the
> > future we
> > have other interesting target like WebAsembly, will want to add
> it and
> > build JS/WEBASM
> >
> > I was unaware of the page so it didn’t get updated with these
> changes to
> > not require SWF artifacts. So it does need updating, but it would
> be best
> > to first make it clear that there are two sets of output.
> >
> > >
> > > Unfortunately, that means that most of the ideas I've read
> while
> > skimming
> > > over this thread so far may not be correct.
> > >
> >
> > I think you have to have in mind that we all was working right
> with our
> > current environment and that the problem comes from try to start
> from
> > scratch. Subsequents builds instructions are simpler since
> requires
> > shorter
> > instructions.
> > You should try to rename your "repository" folder and create a
> new one
> > and
> > try to build with maven to see what you find and if we can
> improve
> > actual
> > findings.
> >
> > Someday I will find time to do that. May not be today. It is, IMO,
> more
> > important for others to understand the goals and how this stuff
> works so it
> > isn't all on me. My understanding of Maven, which I am not an
> expert, is
> > that what is in your local repository shouldn’t matter. Maven goes
> and
> > gets the dependencies you ask for in the pom.xml. The only "trick"
> is how
> > the Mavenizer extension works. That is the only thing that doesn't
> fetch
> > from Maven Central. So renaming or flushing the repository
> "shouldn't"
> > make a difference and someone should figure out why, but only after
> making
> > sure the configurations make sense. Maybe all of the SWC POMs in
> > royale-asjs need a profile that opt-in the SWF artifacts. That
> might be
> > the actual issue. And maybe we set skipTests=false in the compiler
> if not
> > using SWF artifacts via some profile.
> >
> > The key point is that you can't just "get it working for you". We
> have to
> > maintain the two sets of outputs for others.
> >
> > HTH,
> > -Alex
> >
> >
>
> --
> Carlos Rovira
>
> https://nam04.safelinks.protection.outlook.com/?url=http%3A%2F%2Fabout.me%2Fcarlosrovira&data=02%7C01%7Caharui%40adobe.com%7Cb2aa222889e146d467ab08d6dfa95bd1%7Cfa7b1b5a7b34438794aed2c178decee1%7C0%7C0%7C636942315815634189&sdata=idbblu%2BMTCg1mEjTPyymzJ8W1GebB3ksThLI3CzdPr0%3D&reserved=0
>
>
>
