Plastic4: multi-platform installation issues

[Byron]

Hi,

For reasons I explained in another post, I was extremely keen to try Plastic 4, as it was supposed to have a better and more streamlined GUI, besides other features. Basically I did hope that everything would be well with it.

When trying to get Plastic 4 installed, I had to realize that it doesn't make installation easy (if you are anywhere else but a Windows 7 system) - but more on that is to follow in a more structured manner.

General

Please correct me if I am wrong, but it appears that Plastic 4 has now been released officially. However, if you follow the obvious links to the documentation, you will quickly find spots which still refer to the Plastic 3 documentation, as indicated by the screen shots. Maybe I just don't know where to look.

From what I found out so far, Plastic 4 will not read or write into Plastic 3 repositories. This is fair enough, but somewhat unusual considering (even) SVN maintained backward compatibility. If you decide to break it, I would at least expect a clear and easily accessible migration path.

On the download page, there is a green box stating:

Note: if you wish to test Plastic SCM by importing your existing data, please contact support@codicesoftware.com for import details specific to your existing legacy SCM system.

This is about the same message you get once the installation of Plastic 4 finishes on those windows machines lucky enough to run Windows 7. Now, Plastic 3 is a 'legacy' system, and it seems there is no support for it anymore. Migrating existing repositories is enough of a special case for you to make people ask the same question to your support over and over again. Why you wouldn't put up a web site about that topic is something I will never understand, yet it fits well into the general situation that there seems to be no dedicated Plastic 4 documentation online.

Lets go back to your great announcement of Plastic 4. You certainly did a good job hyping it, and I bet that many Plastic 3 users, for one reason or another, badly wanted to have the new version.

A few minutes after the download area showed the new release, I tried to get a copy of all versions. The Linux installer redirected to an invalid URL, apparently it wasn't ready yet. A few minutes later , there was a note at least that indicated the Linux installer would be ready 'in a few hours'. A few hours later, you could still click the broken link, until it was finally removed from the table. The note stating it will be ready 'in a few hours' stayed there - it was misleading and downright wrong, as it would take days until the Linux installer finally appeared.

However, I had two installers to play with, and there certainly were enough 'fun' already.

Installation on Windows XP 64 (SP2)

1. 
   When you try to install the server components, you will see this error message.
   
2. 
   Trying to start the client brings up this error message.
   

About 1.: Apparently the installer just cannot deal with XP 64, as it recommends installing SP3, which simply doesn't exist for XP64. The latest service pack you can obtain is SP2. Thinking further, I thought that maybe the installer would be happy if I just installed SQL Server 4.0 CE directly , but I was just getting the same error message. Why don't you put in a button that says "Proceed Anyway" to allow users to choose a non-default backend for the plastic 4 server, isn't it true that you support that many ?

The only way to get past this point and proceed to 2. is to not install the server components.

Maybe I didn't realize that 64 bit systems are not supported, but from what I find on the download page at least, I wouldn't expect it.

Plastic SCM runs on Windows 7, Windows Server 2008, Vista, Windows 2003 and Windows XP. It requires .NET Framework 2.0 SP2 or higher to run on all Windows platforms. The .NET framework is installed on most new Windows computers. If your system does not have the .NET framework, you can download it from this link:

64 bit systems are not mentioned specifically, which suggests it doesn't matter. Yet the XP installer seems no be misguided on 64 bit versions of windows XP.

About 2.: The client will not start, its as simple as that, and I found no fix for it. Installing .Net 2 to 4 didn't improve a thing. Maybe its just this machine I thought, and went to another XP 64 (SP2) machine which is rather fresh and clean, but its the same issue there. Maybe its possible to test for features/modules that you need in your runtime environment, and catch them more gracefully together with a minimum version that you need. I don't know about .Net, but what can I do more than installing all the versions there are up to version 4 and try to start the client after each one ?

I repeated this with the download of today, and its still the same issue.

My Conclusion: Plastic 4 does not support XP 64 such as Plastic 3 did, and what's even worse, its left to the user to find out.

Windows 7 64

The installer actually succeeded ! When running the server, it used the default back-end which should be SQL 4.0 CE. We managed to make about one commit before the database was "locked", so the server process couldn't access it anymore. This seems to be a typical windows file locking issue, although I didn't expect the server to cause it by itself.

Short after, we just gave up on this as it kept happening over and over again after restarting the server.

OS X 10.7

Its summed up in [url=""%5Danother thread%5B/url%5D. When I first tried it, I had exactly the same problem. Now I tried again, and it seems to be fixed, so I can run the server as well as the client.

The gui likes to freeze if you actually use it - its a known 'inconvenience' from the Linux side - here I assume its related to mono, yet it renders the gui close to being unusable.

Linux

Here it was pretty much the same as on OSX, namely something with mono didn't work so the client couldn't start, and the server failed to start as well.

To be sure, I tried it again today and have to say that the issue seems to be fixed now, allowing me to run the server as well as the client. No trace of it in the fix in change log though !

Its interesting that the --daemon flag used by the installer to launch the server isn't documented when executing the server with –help.

The UI is as instable as the plastic 3 UI, but I am used to that by now, I assume its a mono issue.

What's even worse are the fast-import issues that I am not getting into detail now, but lets state that plastic4 fails to import a repository that plastic 3 exported using the fast-export format. I managed to fix it up manually and import it into git, but even there the folder structure doesn't match the one we see in plastic. Its just 1200 revisions, many branches, and many merges.

Comment on the overall experience

The term 'desaster' probably fits rather well here. I wasted so much of my time to try broken installers, they caused a lot of grief. Plastic 4 was announced as the great thing, but it was nothing more than a whole lot of marketing buzz. The marketing dept. tries to make it something that it simply is not, making the disappointment even larger! Plastic 4 is just not ready, it seems to be totally beta, and I do wonder if anybody tests it properly on non-windows systems. The GUIs are just instable there. (Mono may play a huge role here as this is not the first mono application which tends to crash on non-windows platforms).

There is no documentation, the existing one refers to plastic 3, and honestly, you are better than that !

As a developer, what I see here is a top-down decision saying that the release has to be made, no matter what. Even though you made it on time, the damage it is causing costs your company so much credibility and users. I feel sorry for the developers who told management that this is a bad idea, and whoew2q were not heard.

For my part, I have had enough, and we are pro-actively looking for a way out. This is sad though, as a lot of investment goes down the drain. Plastic SCM was promising GUI wise, but it seems to loose it.

If I was diplomatic or a manager, I would now delete this post and start over again, rewriting it in a form which merely states the issues and kindly asks for fixes. However, I believe its best not to spare all the emotion that went into it to show exactly how I feel. Plastic SCM could be good, but it just isn't, and unfortunately its advertised differently which doesn't make anything better.

I wish you good luck in fixing this release and documentation, there is still a lot of work to do.

Also I am very sorry to have felt the need to write a post in this tone as I believe all of you did their best to deliver the best product they could within the given constraints.

Sebastian

[psantosl]

Hi Sebastien,

First of all, thanks a lot for the detailed report.

I'm far from happy to read this. It has been a lot of work to come up with 4.0, really a daunting task for a team our side.

Let me go through your post trying to answer you.

We do care, we do care about our users.

Let's go:

For reasons I explained in another post, I was extremely keen to try Plastic 4, as it was supposed to have a better and more streamlined GUI, besides other features. Basically I did hope that everything would be well with it.

When trying to get Plastic 4 installed, I had to realize that it doesn't make installation easy (if you are anywhere else but a Windows 7 system) - but more on that is to follow in a more structured manner.

Sebastien, this is the list of operating systems we use to run our entire test suite on:

- Windows 2000, XP, Vista, 7, 2003Server, 2008 server. I don't remember now which ones we run on 64 bits but I can ask.

Remember: 2k and XP are no longer supported by Microsoft. We do our best to support them.

General

Please correct me if I am wrong, but it appears that Plastic 4 has now been released officially. However, if you follow the obvious links to the documentation, you will quickly find spots which still refer to the Plastic 3 documentation, as indicated by the screen shots. Maybe I just don't know where to look.

You're right. We updated the distributed guide, the rest of the guides will be published this week and next, but unfortunately aren't there now.

It is not an excuse, it is our responsibility, but we simply didn't have the time to complete everything.

From what I found out so far, Plastic 4 will not read or write into Plastic 3 repositories. This is fair enough, but somewhat unusual considering (even) SVN maintained backward compatibility. If you decide to break it, I would at least expect a clear and easily accessible migration path.

Well, we made a big conceptual change in 4.0, basically now each "tree" is static (sort of like git) instead of dynamic as it used to be. It is just a small change on the database but the migration requires loading each changeset (its selector) and making some adjustments, so after 5 years, this is the first non-upgradeable version we release.

The migration path was requested to be published one week ago, I'll double check why it is not.

On the download page, there is a green box stating:

This is about the same message you get once the installation of Plastic 4 finishes on those windows machines lucky enough to run Windows 7.

Well, I understand your frustration, but yes, it does run on more machines than w7.

As a side note: XP is 11 years old.

Now, Plastic 3 is a 'legacy' system, and it seems there is no support for it anymore.

Well, we do support 3.0. Who told you we don't?

Most of our customers are in 3.0 yet... we do actively support them. In fact the latest 3.0 release is not more than a month old or so!

Migrating existing repositories is enough of a special case for you to make

people ask the same question to your support over and over again. Why you

wouldn't put up a web site about that topic is something I will never understand,

yet it fits well into the general situation that there seems to be no dedicated

Plastic 4 documentation online.

Well, I wrote the text myself to get it published like one week ago. I'll check what went wrong.

Lets go back to your great announcement of Plastic 4. You

certainly did a good job hyping it, and I bet that many Plastic 3

users, for one reason or another, badly wanted to have the new version.

Well, yes, many teams, sure. The big ones are in direct contact with us through support.

Most of paying customers already contacted us and got an answer.

Which doesn't mean it shouldn't be doc online.

A few minutes after the download area showed the new release, I tried to get a copy of all versions. The Linux installer redirected to an invalid URL, apparently it wasn't ready yet. A few minutes later , there was a note at least that indicated the Linux installer would be ready 'in a few hours'. A few hours later, you could still click the broken link, until it was finally removed from the table. The note stating it will be ready 'in a few hours' stayed there - it was misleading and downright wrong, as it would take days until the Linux installer finally appeared.

However, I had two installers to play with, and there certainly were enough 'fun' already.

pre-4.0 on Linux has been up and running on some huge customers for months.

I personally decided to hold the release a few days because I wanted the team to double check a few issues we found.

Installation on Windows XP 64 (SP2)

When you try to install the server components, you will see this error message.

As far as I see this is an info message, not an error, is it?

Trying to start the client brings up this error message.

That's obviously an error. We run tons of setups and we found it only a few times, we're trying to find out the issue, but it is quite uncommon and normally only happens to people trying to connect to 3.0 with a 4.0 client. Can be the case?

[psantosl]

(continued, the damn forum software doesn't let me put as many quotes as I'd like... )

About 1: Apparently the installer just cannot deal with XP 64, as it recommends installing SP3, which simply doesn't exist for XP64. The latest service pack you can obtain is SP2. Thinking further, I thought that maybe the installer would be happy if I just installed SQL Server 4.0 CE directly , but I was just getting the same error message. Why don't you put in a button that says "Proceed Anyway" to allow users to choose a non-default backend for the plastic 4 server, isn't it true that you support that many ?

There's a switch to skip all checks.

It is not there by default, that's true, on the GUI.

We must getting better documented, that's right and I'm sorry again for the inconveniences.

The only way to get past this point and proceed to 2. is to not install the server components.

Maybe I didn't realize that 64 bit systems are not supported, but from what I find on the download page at least, I wouldn't expect it.

64 bit systems are not mentioned specifically, which suggests it doesn't matter. Yet the XP installer seems no be misguided on 64 bit versions of windows XP.

That's true Sebastien: when we upgraded to 4.0 and changed the installation requirements, we run our tests on the "testing mini-cluster" on the available xp machines. And certainly we did not cover 64bits.

Again, we're sorry for this.

The only point to mention is: xp 64 is not very common, again just an excuse. We must directly say it is not supported or iron out the installation issues.

About 2.: The client will not start, its as simple as that, and I found no fix for it. Installing .Net 2 to 4 didn't improve a thing. Maybe its just this machine I thought, and went to another XP 64 (SP2) machine which is rather fresh and clean, but its the same issue there. Maybe its possible to test for features/modules that you need in your runtime environment, and catch them more gracefully together with a minimum version that you need. I don't know about .Net, but what can I do more than installing all the versions there are up to version 4 and try to start the client after each one ?

I'm not aware of this. Of course these things happen, but it is not the usual case. It won't make you feel better because YOU are experiencing all this, but we talk to hundreds of users and fortunately this is pretty uncommon. I never seen a client "non-starting". It always gives a reason why it can't run.

I repeated this with the download of today, and its still the same issue.

My Conclusion: Plastic 4 does not support XP 64 such as Plastic 3 did, and what's even worse, its left to the user to find out.

There are two big changes on 4.0 regarding this:

- the server requires sqlserver ce by default

- the client is built on .net 2.0 -> meaning it will run as a 64 bits process by default.

Windows 7 64

The installer actually succeeded ! When running the server, it used the default back-end which should be SQL 4.0 CE. We managed to make about one commit before the database was "locked", so the server process couldn't access it anymore. This seems to be a typical windows file locking issue, although I didn't expect the server to cause it by itself.

Short after, we just gave up on this as it kept happening over and over again after restarting the server.

God! Sebastien! We've been using this thing for 12 months down here, we even run load tests on it (100 client (actual machines automated by the test system) against a single 4.0 server). We didn't see "a lock". So, not sure why this is happening.

We're always available to help our users, so instead of going through all the frustration, please come to us and we'll try to help.

OS X 10.7

Its summed up in [url=""%5Danother thread%5B/url%5D. When I first tried it, I had exactly the same problem. Now I tried again, and it seems to be fixed, so I can run the server as well as the client.

The gui likes to freeze if you actually use it - its a known 'inconvenience' from the Linux side - here I assume its related to mono, yet it renders the gui close to being unusable.

I don't have much to say here. We have users working on mac and linux, but you must have reasons to say it is "unusable".

Linux

Here it was pretty much the same as on OSX, namely something with mono didn't work so the client couldn't start, and the server failed to start as well.

To be sure, I tried it again today and have to say that the issue seems to be fixed now, allowing me to run the server as well as the client. No trace of it in the fix in change log though !

Its interesting that the --daemon flag used by the installer to launch the server isn't documented when executing the server with –help.

The UI is as instable as the plastic 3 UI, but I am used to that by now, I assume its a mono issue.

Mono is pretty stable on the server side, so it can be a config thing on the client, or a gui bug. An specific scenario would certainly help.

What's even worse are the fast-import issues that I am not getting into detail now, but lets state that plastic4 fails to import a repository that plastic 3 exported using the fast-export format. I managed to fix it up manually and import it into git, but even there the folder structure doesn't match the one we see in plastic. Its just 1200 revisions, many branches, and many merges.

Well, 3.0 fast-export is included on the release but never announced for a reason. Branches inherited from "last" on 3.0 (normally created by mistake of lack of understanding by users) are not easy to introduce on the more restrictive 4.0 schema. That's why we didn't announce the migration path yet, because we're still working on it.

Comment on the overall experience

The term 'desaster' probably fits rather well here. I wasted so much of my time to try broken installers, they caused a lot of grief. Plastic 4 was announced as the great thing, but it was nothing more than a whole lot of marketing buzz. The marketing dept. tries to make it something that it simply is not, making the disappointment even larger! Plastic 4 is just not ready, it seems to be totally beta, and I do wonder if anybody tests it properly on non-windows systems. The GUIs are just instable there. (Mono may play a huge role here as this is not the first mono application which tends to crash on non-windows platforms).

Well, it is hard for me to read that the new branch explorer, new merge system (steps ahead of any other out there), new distributed stuff, the sync view, xlinks, move detection and many mor is "a whole lot of marketing buzz".

Of course I respect your point and it will help us to create a better product.

There is no documentation, the existing one refers to plastic 3, and honestly, you are better than that !

As a developer, what I see here is a top-down decision saying that the release has to be made, no matter what. Even though you made it on time, the damage it is causing costs your company so much credibility and users. I feel sorry for the developers who told management that this is a bad idea, and whoew2q were not heard.

I made the decision, so I'm the one to blame. 4.0 is not perfect, sure, but is is way much better than 3.0, and better than the other SCMs out there. We'll continue working hard to come up with more and more polished releases, but I can assure you your statement is far from accurate. I respect your point anyway. But, please, remember most of the frustration came from a set up made on XP 64, which, as great as it is, is an old and tough to support piece.

We certainly have to improve linux and mac gui support, although most of our users there tend to use the cli.

Ironically, our biggest servers run on Linux, with excellent results.

But admittedly, we focus less on linux gui than windows, basically because most of our users are windows ones.

[psantosl]

(and more)

For my part, I have had enough, and we are pro-actively looking for a way out. This is sad though, as a lot of investment goes down the drain. Plastic SCM was promising GUI wise, but it seems to loose it.

Well, as much as it hurts me, we can't be the tool for everyone.

We are not a tool for open source projects because git/github is most oriented towards it than we're.

We're a tool for commercial teams.

We're completely dedicated to our customers, and that's why you can request a bugfix and getting it in a matter of hours. And you know better than me that this is not the case for the tool you're looking to move to.

4.0 contains a ton of great features. Most of them are answers to customers requests. We do listen, and we do react. Please, consider it again before abandoning.

One of the things I really admire from github is that each time they announce a server crash or downtime, their fans applaud and encourage them to try harder.

We fail to install on an OS unsupported by its manufacturer and we get a great developer blaming us publicly.

If you put effort on Plastic, please, don't give up. We'll work hard to help you, we will definitely come up with a better and better product. But please, be fair and don't subestimate the whole set of advantages we provide.

We need people like you using plastic, helping us come up with a better and better product. Developers who know what they want and never lower their expectations. We ask for a little of patience, and I certainly can tell you that we can make it work on xp 64 for you with just a little bit of configuration.

If I was diplomatic or a manager, I would now delete this post and start over again, rewriting it in a form which merely states the issues and kindly asks for fixes. However, I believe its best not to spare all the emotion that went into it to show exactly how I feel. Plastic SCM could be good, but it just isn't, and unfortunately its advertised differently which doesn't make anything better.

I wish you good luck in fixing this release and documentation, there is still a lot of work to do.

Also I am very sorry to have felt the need to write a post in this tone as I believe all of you did their best to deliver the best product they could within the given constraints.

Sebastian

Thanks for being so direct. It will help us to become better.

[Byron]

Once again, thanks for your elaborate reply. This really gives me hope that we will get our issues solved after all. In fact I planned to reply much sooner, but decided I wait until I have plastic 4 running with our converted repositories to have a reason to cheer and provide a positive testimonial.

WinXP64ServerComponents.png

As far as I see this is an info message, not an error, is it?

Yes, its a message indeed. However, you may either go back to where you came from (and hitting next triggers the very same message box), or you hit yes and open a website, and then go back to where you came from. In other words, without using the commandline invocation workaround, you will not get past this requirement. In fact, it just doesn't correctly verify the prerequesites, as I installed everything it could ever need, and managed to run the server once I made the installation pass.

That's obviously an error. We run tons of setups and we found it only a few times, we're trying to find out the issue, but it is quite uncommon and normally only happens to people trying to connect to 3.0 with a 4.0 client. Can be the case?

Yes, this was the case, and your hint helped to solve the problem ! Thank you for that !

God! Sebastien! We've been using this thing for 12 months down here, we even run load tests on it (100 client (actual machines automated by the test system) against a single 4.0 server). We didn't see "a lock". So, not sure why this is happening.

We're always available to help our users, so instead of going through all the frustration, please come to us and we'll try to help.

Yes, I will do this from now on, promised ! Your support is excellent, there is nothing else to say about that. By now I managed to use it on Win 7 without this lock issue - and I have no idea why this happend the first time. A reinstall did it, but by that time it must have been a newer version as well. Anyhow, lets forget about this.

I don't have much to say here. We have users working on mac and linux, but you must have reasons to say it is "unusable".

It crashed a lot, and its still doing so occasionally. However, I use the mono version you supply for plastic 4 in plastic 3 (because I have to), and it seems to have improved the overall stability.

Mono is pretty stable on the server side, so it can be a config thing on the client, or a gui bug. An specific scenario would certainly help.

Once I use Plastic 4 in production, I provide all information I can find to help making it more stable.

We certainly have to improve linux and mac gui support, although most of our users there tend to use the cli.

That's what I thought as well at first, using the cli could probably help. However, I find it uncomfortable especially compared to what I am used to. I wouldn't say git has the best cli ever, but it certainly makes the common things easy.

From my plastic 4 tests, I must absolutely admit that I can get used to the new branch explorer and the distributed workflow which has become usable, so I don't even want to use the cli for daily use. Maybe just me though .

We need people like you using plastic, helping us come up with a better and better product. Developers who know what they want and never lower their expectations. We ask for a little of patience, and I certainly can tell you that we can make it work on xp 64 for you with just a little bit of configuration.

To sum it up, using the command line options of the installer allows to install the server on xp 64 as well, and run it successfully.

Its certainly not my decision to which SCM we would change if we dropped plastic, and as there are just no alternative providing a gui on a comparable level as the one in plastic 4, we decided to drop plastic 3 for its successor.

Your ongoing support and professional handling of your community make me believe that - how frustrating our issues might be - we will get it solved with your help after all !

Your help I will need indeed to finalize the move to plastic 4, but I will put all that into a new topic.

I keep going !

Sebastian

[manu]

Hello Sebastian,

as I said before thanks for returning with your feedback.

Your help I will need indeed to finalize the move to plastic 4, but I will put all that into a new topic.

I keep going !

It's very important contacting us before getting frustrated.

You know you can ask everything you want, we are here to help you.

Regards,

Manu.