Hi Dave, Maven was easily the most asked for feature for Flex in the Adobe days. The number of votes for it outnumbered the second place wish-list item by a factor of 10, IIRC.
So we (mostly Christofer Dutz) made Maven work for FlexJS/Royale. Once you buy into Maven, you want to rely on its prepare, build, packaging features. Meanwhile many other Flex/Royale customers seem to still want to use Ant/Command-line. So we have Ant scripts as well. Maven doesn't like Ant, so we "must" use Ant to automate our release process and have it call Maven as needed. It might be better to try to break it up into manual steps, but IMO, there's be more confusion about what the words in the run book actually mean, folks would miss important steps, and the kinds of problems we're seeing would still happen and the interventions would be the same. IOW, we have a complex set of steps to create a release (yes, I know, it pales compared to AOO). The Ant script does break it up into pieces, but Maven does large uninterruptable chunks and doesn't recover well. IMO, the point of having others be RM now is to find the bugs in the scripts so that future RMs won't be frightened away by the complexity, and we are finding that folks have broken configurations like funky uploading or forgetting their GPG passwords and it is all part of the learning process. I am purposefully not volunteering to be the RM because I want to make sure at least one other person on the PMC can truly understand the steps. Right now, I'm not sure the level of understanding is there and folks are blindly running the scripts hoping it will work and then have to wait for me to help them. That needs to change as well, but it is also not an efficient use of my time to guess where things are going to break and write up a huge document that tries to capture everything. We have a wiki doc that we try to capture stuff as we run into it. Ant Scripts "should" be relatively straightforward to decipher. Maven steps are documented in the Maven doc. IMO, every PMC member needs to be relatively good at Ant and Maven, not to mention Java, AS3 and MXML. If Royale grows to have more users, the committers and PMC members are going to have to get good at debugging any of our code. It can't keep waiting on me. I only originally authored a tiny fraction of the code. A significant amount of the time, I am trying to figure out someone else's code as well. To me, debugging other people's code is an important capability for an Open Source developer. We are encouraging others to change our code. We have to get efficient at debugging things we didn't write. Now the repos are in an uncertain state and I guess I will have to spend some time bringing back to where it needs to be. But there isn't much point in that until Carlos can find a workflow that does not change tags. Otherwise, every commit from Carlos will put us back into a bad state. I think if Piotr starts over from an empty folder and remembers his GPG password, there is a good chance he will get through it. And if not, we simply have to continue to make the process more robust for every future RM. My 2 cents, -Alex On 10/11/18, 6:07 PM, "Dave Fisher" <dave2w...@comcast.net> wrote: Hey Guys, I was discussing with a friend on another project Any vs. Maven. Ant is nice and procedural while Maven is declarative. Lots of projects use Ant like Tomcat. I suspect that here in Royale we don’t have clear understanding about what to change in build.xml and pom.xml when we branch or tag. If these changes aren’t exceedingly simple then maybe there is a need for a configure script that makes needed edits. Like - sed -e config.sed build-proto.xml > build.xml sed -e config.sed pom-proto.xml > pom.xml I’ve been wondering about this release for awhile. Is the release a branch that is tagged? If it isn’t then maybe that is an issue. Why would develop interfere? Maybe the process needs to made into steps. There needs to be a prepare script, a build script, and a package script. I’m finding that no one can seem to be a successful RM disappointing. Regards, Dave Sent from my iPhone > On Oct 11, 2018, at 5:12 PM, Carlos Rovira <carlosrov...@apache.org> wrote: > > Alex, > > read the other thread where I just explain why you are not seeing the build > fail. The problem is the release process that are making us think all is > ok. but Maven is not ok. Only ANT since is not affected by release process. > As I said in the other thread. We can't left a unfinished release process > not reverted for several days since. > > Hope it makes it more clear. I'm closing for today > > Carlos > > > > El vie., 12 oct. 2018 a las 1:59, Alex Harui (<aha...@adobe.com.invalid>) > escribió: > >> Carlos, >> >> Why is it that last night the builds worked for me without making any >> changes and the Maven build on builds.a.o was working? Now all of these >> changes and I have no idea if they make a difference or not. >> >> When something is not working for you but working for others, making >> changes that affect everyone is probably not a good idea unless you really >> understand why it isn't working for you. >> >> -Alex >> >> On 10/11/18, 4:56 PM, "Carlos Rovira" <carlosrov...@apache.org> wrote: >> >> Hi, >> >> In each repo I created a branch "develop-fix" to try to get all compile >> again, and trying not to interfiere with release process (although I >> suppose this will be reverted, in that case maybe we only need to merge >> this fix branches). >> >> compiler and typedefs are easy: just revert the two commits to prepare >> relase >> >> framework repo is not as direct: >> >> - MXRoyale fails since it needs the mxml-2009-manifest (see this >> commit 1efd8fb1ba6d6894a542148e5dd8210830c0b097), I tried to add to >> pom.xml >> but there's a class that is not found "DesignLayer" that I commented >> in a >> copy of the manifest in the MXRoyale project for maven. Sure Alex will >> find >> a better solution for this. >> >> With this framework compiles ok. >> >> -Then Examples: Some examples are failing due to fx:Array >> >> like: >> >> >> /Users/carlosrovira/Dev/Royale/Source/royale-asjs/examples/royale/AccordionExample/src/main/royale/AccordionExample.mxml(33): >> col: 25 Error: In initializer for 'fx:Array' multiple initializer >> values >> are not permitted for target type '*'. >> >> >> <js:Container id="panel1"> >> >> ^ >> >> >> >> /Users/carlosrovira/Dev/Royale/Source/royale-asjs/examples/royale/AccordionExample/src/main/royale/AccordionExample.mxml(44): >> col: 25 Error: In initializer for 'fx:Array' multiple initializer >> values >> are not permitted for target type '*'. >> >> >> <js:Container id="panel2"> >> >> ^ >> >> or: >> >> >> /Users/carlosrovira/Dev/Royale/Source/royale-asjs/examples/royale/RoyaleStore/src/main/royale/SupportView.mxml(73): >> col: 33 Error: In initializer for 'fx:Array' multiple initializer >> values >> are not permitted for target type '*'. >> >> >> <fx:String>California</fx:String> >> >> ^ >> >> >> >> /Users/carlosrovira/Dev/Royale/Source/royale-asjs/examples/royale/RoyaleStore/src/main/royale/SupportView.mxml(74): >> col: 33 Error: In initializer for 'fx:Array' multiple initializer >> values >> are not permitted for target type '*'. >> >> >> <fx:String>Nevada</fx:String> >> >> ^ >> >> Maybe due to the way I add the manifest before? >> >> >> - >> >> -- >> Carlos Rovira >> >> https://na01.safelinks.protection.outlook.com/?url=http%3A%2F%2Fabout.me%2Fcarlosrovira&data=02%7C01%7Caharui%40adobe.com%7Ca5c2ef11cc8c4cea6b1808d62fdf20c8%7Cfa7b1b5a7b34438794aed2c178decee1%7C0%7C0%7C636749032708199818&sdata=aCWLbPp9UmhgOhphoiaueTaZ9PDwxoPN1k3Bv278AAg%3D&reserved=0 >> >> >> > > -- > Carlos Rovira > https://na01.safelinks.protection.outlook.com/?url=http%3A%2F%2Fabout.me%2Fcarlosrovira&data=02%7C01%7Caharui%40adobe.com%7Ca5c2ef11cc8c4cea6b1808d62fdf20c8%7Cfa7b1b5a7b34438794aed2c178decee1%7C0%7C0%7C636749032708209819&sdata=721qTEbEIymseUAC5AXfMkdqIoGDNcDR%2F9S4w%2BJS92Y%3D&reserved=0