Not Fit For Release

I'm going to rant... right... NOW:

I really hope that RC1 was a joke... there are (in my mind) two HUGE blockers that should prevent any more builds from being called a release candidate:

  • The J2EE support is not exposed as an API. This means all the nice J2EE module support is completely useless to anyone who wants to use Resin or Orion. There is almost zero gain in using 4.0's J2EE module support over 3.0's basic web app support. Things like the WebRoot must be exposed in the API so that the Resin and Orion plugins (Tomcat is too crappy to know what to do with this) can generate correct deployment/config files.


  • Jikes compile support is totally broken. I've seen previous threads complaining this is a jikes bug and what-not... but the fact of the matter is that javac (make|rebuild|compile) all work, but jikes rarely does. This is an unacceptable regression! Anyone who has a large project can attest to the fact that jikes makes a huge difference in how long we wait around for a clean build. PLEASE PLEASE PLEASE get this fixed.


Pat

28 comments
Comment actions Permalink

Well I don't know about that particular issues, but I can only second that
this is really not to be called
Release Candidate.
The two most important sections for me are:

- There are still very serve memory leaks in there: see the following very
high voted bugs: 7069, 16948
- There are serveral big bugs with core refactoring features, with new ones
poping up every day.


0
Comment actions Permalink

Mike The Grey wrote:

Well I don't know about that particular issues, but I can only second that
this is really not to be called
Release Candidate.
The two most important sections for me are:

- There are still very serve memory leaks in there: see the following very
high voted bugs: 7069, 16948


16948 seems to be fixed - see discussion

- There are serveral big bugs with core refactoring features, with new
ones poping up every day.

Which are they?

Friendly,
Dmitry

--
Dmitry Lomov
Software Developer
JetBrains Inc.
http://www.jetbrains.com
"Develop with pleasure!"

0
Comment actions Permalink

Perhaps in an attempt to gain closure to the problem set, and to get the remaining "hot" issues identified, our JetBrains friends called this the "release candidate" in order to get our attention.

Besides, I would guess that for the majority of users (like me), it works just fine. Release it for us (so jetbrains can get some money -- just got my 4.0 Purchase Order processed), and then fix subsequent issues in 4.01 or 4.02 releases.

They do, afterall, have to eat something other then snow :)

Mike

p.s. Although it would be interesting to see if the St Petersburgh winter is as bad as it is here in Rochester New York....We've had 3 straight weeks of snow...a new record.

0
Comment actions Permalink

On 2004/01/27 17:44, Pat Lightbody wrote:

  • Jikes compile support is totally broken. I've seen previous threads

complaining this is a jikes bug and what-not... but the fact of the
matter is that javac (make|rebuild|compile) all work, but jikes
rarely does. This is an unacceptable regression! Anyone who has a
large project can attest to the fact that jikes makes a huge
difference in how long we wait around for a clean build. PLEASE
PLEASE PLEASE get this fixed.


I have attached test cases to the following SCRs
http://www.intellij.net/tracker/idea/viewSCR?publicId=24842 (Compiler hangs)
http://www.intellij.net/tracker/idea/viewSCR?publicId=23451 (jikes
support broken)

I'm hoping that helps a bit so that those bugs may be fixed before
releasing 4.0

Bas

0
Comment actions Permalink

I don't think it makes sense to list them. I know that you guys are working
hard to get them fixed.
I just don't feel very comfortable knowing that the refactorings i use might
or might not work.
The whole thing about core features like extract method is that I don't want
to bother with that
myself and rely on the IDE doing it right.

Don't get me wrong the latest versions are getting in the right direction
and it is possible
to just work with the EAP, but it just isn't as stable as it should be.

If something is called a Release Candidate it suggest that it might be
relabeled (the same build) to Final Release
if it has no more critical bugs.


0
Comment actions Permalink

Mike The Grey wrote:

I don't think it makes sense to list them. I know that you guys are
working hard to get them fixed.


It does, I guess. I am not aware of any bug in refactoring that stops us
from shipping 4.0. If you do, please let me know.

Friendly,
Dmitry
--
Dmitry Lomov
Software Developer
JetBrains Inc.
http://www.jetbrains.com
"Develop with pleasure!"

0
Comment actions Permalink

I have to concur that the overall feeling from IDEA is that it is not
release-ready. You may ask about particular bugs but this is getting
more about perception - too fragile, too much special knowledge to get
things done right.
As an experiment you may work with some user to write an enthusiastic
message Subject: IDEA rocks or so and have some areas mentioned like
Ant support, EJBs, WebApps (JSPs and HTML), refactoring or VCS support.
Then watch the responses - if most of them are positive - you are ready
to ship. If you see mostly complaints ("You say it rocks? Why? Did you
try <....>?") you know what to do.

If things are too subtle, too fragile, you may not get as many SCRs just
because many of us are afraid of walking on glass when some work has to
be done (one problem and I back off, no time to reconstruct).

r.

0
Comment actions Permalink

http://www.intellij.net/tracker/idea/viewSCR?publicId=27098

Talks about crippling instability in the debugger. Unless someone can offer a satisfactory explanation for this behavior (like stupid user error on my part), I would expect this prevent release.

0
Comment actions Permalink

This is exactly what I mean. if the release would be stable over all, there
wouldn't be so many reports for every single
build:

1113 119 bugs
1117: 49 bugs
1120: 188 bugs and exception reports
1122: 42 bugs

and that all in a couple of days.


0
Comment actions Permalink

what about 19351. Idea tells me that my code is wrong and then when it fixes it for me it won't compile

0
Comment actions Permalink

Yep, it certainly got me to panic and start filing bugs about all the things that annoy me!

0
Comment actions Permalink

see http://www.intellij.net/tracker/idea/viewSCR?publicId=27264

they are the same I guess

"charles decroes" <spam@decroes.com> wrote in message
news:26927674.1075302321041.JavaMail.itn@is.intellij.net...

what about 19351. Idea tells me that my code is wrong and then when it

fixes it for me it won't compile


0
Comment actions Permalink

  • Jikes compile support is totally broken. I've seen previous threads complaining this is a jikes bug and what-not... but the

fact of the matter is that javac

Well' this is indeed jikes's bug. See what it prints to console when invoked from the command line:

Assertion failed: false && "LoadVariable bad kind", file ../../jikes-1.18/src/bytecode.cpp, line 6350
This application has requested the Runtime to terminate it in an unusual way.
Please contact the application's support team for more information.

The olny problem from IDEA's side is that it waits for the jikes process that never finishes. We've found the workaround for that
and now IDEA won't hang waiting. Will be available in next public build

--

Best regards,
Eugene Zhuravlev
Software Developer
JetBrains Inc
http://www.jetbrains.com
"Develop with pleasure!"



0
Comment actions Permalink

My biggest problem with the current release is with cvs. It does not
support local access, and the ssh implementation does not handle
password access.

Add either, and I could work with my local repository without a problem.
Without either one, though, I have to open up security holes I am not
comfortable with.

pserver gives me a big old security hole that I am not willing to
accept. Yeah, it is arguable how damaging that hole is, but I trust
very little going over my network unless it is ssh-encrypted.

If I use local access, NO port on my machine is open. Nobody, as long
as the OS is well implemented, can reach ANY local resources.

ssh, allowing passwords, is somewhat of a risk, but one I can accept.
ssh has its uses for other purposes.

ssh, requiring private key files, will NOT work with sourceforge, and
requires me to have a security hole I would rather not set up. That is
two strikes. (Setting up ssh-agent seems to be a bit of overkill, and I
am not eager to climb that learning curve just for one IDE.)

This is a stopship, IMO, as the cvs features are a big part of the
reason for this release. I can agree that they would be nice, if only
they worked on my local projects OR my sourceforge ones. The fix would
be to either implement local access, or to implement ssh in such a way
that a passphrase is acceptable.

My opinions only, of course

Scott
Java, Cocoa, WebObjects, and database consulting

0
Comment actions Permalink

I have spotted one;

You have two modules. With forms in each module. When moving the form from on module to the other, or even moving a form around in the package hierarchy will eventually break the code. The root-references in the form is what is broken...

Familiar?

Regards
Chris

0
Comment actions Permalink

Yeah, I can't even run the installer without it locking up at least once while copying the JRE or some html files.

0
Comment actions Permalink

Eugene,
I don't ever see that error... in my situations I am getting reports that a particular class or package does not exist, even though it very much does. What bothers me is that this works in 3.0 (same module/project) and not in 4.0 and it sounds like you 4.0 is going to ship without this being fixed.

Is there any information I could give you to help you understand the problem I am seeing better? I do see that jikes.exe is staying alive in the background (clicking stop after the error message doens't kill it, I have to kill the process manually). The thing is that jikes is saying it can't find certain packages, but those packages exist in the same source directory as other classes that are compiling.

Is there a way I can find out what external process you are running (ie: "jikes.exe ...") along with the command line options given? That way I can run it by hand and try to understand better where the failures are happening.

Pat

PS: what about J2EE API support? Having web modules is fantastic -- I LOVE this feature -- but it is almost entirely useless without allowing that new information stored in the module to be extracted via the API and then used to construct files such as resin.xml or Orion's server.xml.

0
Comment actions Permalink

The one thing that is weird is how a certain bug gets fixed and then comes
back a few builds later..

In 1126 these bugs weren't causing me problems and now are again in 1126.
Both are extremely annoying and basically make the build unusable for me at
all.

1) Editing a JSP crashses IntelliJ:
http://www.intellij.net/tracker/idea/viewSCR?publicId=28133

2) Compiling any class crashes IntelliJ:
http://www.intellij.net/tracker/idea/viewSCR?publicId=28109


--Grant


"Pat Lightbody" <plightbo@hotmail.com> wrote in message
news:16118839.1075221875651.JavaMail.itn@is.intellij.net...

I'm going to rant... right... NOW:

>

I really hope that RC1 was a joke... there are (in my mind) two HUGE

blockers that should prevent any more builds from being called a release
candidate:
>

  • The J2EE support is not exposed as an API. This means all the nice J2EE

module support is completely useless to anyone who wants to use Resin or
Orion. There is almost zero gain in using 4.0's J2EE module support over
3.0's basic web app support. Things like the WebRoot must be exposed in
the API so that the Resin and Orion plugins (Tomcat is too crappy to know
what to do with this) can generate correct deployment/config files.
>

  • Jikes compile support is totally broken. I've seen previous threads

complaining this is a jikes bug and what-not... but the fact of the matter
is that javac (make|rebuild|compile) all work, but jikes rarely does. This
is an unacceptable regression! Anyone who has a large project can attest to
the fact that jikes makes a huge difference in how long we wait around for a
clean build. PLEASE PLEASE PLEASE get this fixed.
>

Pat



0
Comment actions Permalink

On Wed, 28 Jan 2004 12:25:26 -0800, Scott Ellsworth wrote:

My biggest problem with the current release is with cvs. It does not
support local access, and the ssh implementation does not handle password
access.


Just run a local pserver. As for SSH and passwords, worked for me just
fine....


0
Comment actions Permalink

Passwordless SSH key files do work with SourceForge. I think you set it up in your user preferences, or maybe project preferences.

0
Comment actions Permalink

Chris Patrick Schreiner wrote:

I have spotted one;

You have two modules. With forms in each module. When moving the form from
on module to the other, or even moving a form around in the package
hierarchy will eventually break the code. The root-references in the form
is what is broken...

Familiar?


No. Not reported ever. In fact, I am afraid I failed to understand your
description. Please file an SCR for this.


Friendly,
Dmitry
--
Dmitry Lomov
Software Developer
JetBrains Inc.
http://www.jetbrains.com
"Develop with pleasure!"

0
Comment actions Permalink

Yeah, I can't even run the installer without it locking up at least once

while copying the JRE or some html files.

It's definitely a stopship and will be fixed eventually. We just haven't
found what's the problem.

--
Valentin Kipiatkov
Chief Scientist, Vice President of Product Development
JetBrains, Inc
http://www.jetbrains.com
"Develop with pleasure!"

"Ian Zabel" <ian@ezabel.com> wrote in message
news:12556900.1075324760899.JavaMail.itn@is.intellij.net...

Yeah, I can't even run the installer without it locking up at least once

while copying the JRE or some html files.


0
Comment actions Permalink

Grant Gochnauer wrote:

The one thing that is weird is how a certain bug gets fixed and then comes
back a few builds later..

In 1126 these bugs weren't causing me problems and now are again in 1126.
Both are extremely annoying and basically make the build unusable for me at
all.

1) Editing a JSP crashses IntelliJ:
http://www.intellij.net/tracker/idea/viewSCR?publicId=28133


The problem is inside your TEI and I think it's repeatable in all IDEA
versions supporting JSP TEI. This is not the reason for IDEA to fall but
it is not a show stopper. Anyway we've fixed one in 1129.

IK
--
Igor Kuralenok
Developer
JetBrains Inc.
http://www.jetbrains.com
"Develop with pleasure!"

0
Comment actions Permalink

Great. I saw that it was fixed... Thank you so much! :)

Its curious why it is having problems finding the class when it is in the
import, for the class, compiles fine and works on the web app just fine. Why
would intelliJ be unable to find that class?

--Grant


"Igor Kuralenok (JetBrains)" <ik@intellij.com> wrote in message
news:bvb87v$s1j$1@is.intellij.net...

Grant Gochnauer wrote:

The one thing that is weird is how a certain bug gets fixed and then

comes

back a few builds later..

>

In 1126 these bugs weren't causing me problems and now are again in

1126.

Both are extremely annoying and basically make the build unusable for me

at

all.

>

1) Editing a JSP crashses IntelliJ:
http://www.intellij.net/tracker/idea/viewSCR?publicId=28133

>

The problem is inside your TEI and I think it's repeatable in all IDEA
versions supporting JSP TEI. This is not the reason for IDEA to fall but
it is not a show stopper. Anyway we've fixed one in 1129.

>

IK
--
Igor Kuralenok
Developer
JetBrains Inc.
http://www.jetbrains.com
"Develop with pleasure!"

>


0
Comment actions Permalink

Grant Gochnauer wrote:

Great. I saw that it was fixed... Thank you so much! :)

Its curious why it is having problems finding the class when it is in the
import, for the class, compiles fine and works on the web app just fine. Why
would intelliJ be unable to find that class?

Good question :). I think the library containing missing class is not
included in deployment. This condition is relaxed in 1129+. Now TEI
classes is just to be resolved in module scope (in output or libraries).

IK

--
Igor Kuralenok
Developer
JetBrains Inc.
http://www.jetbrains.com
"Develop with pleasure!"

0
Comment actions Permalink

Perfect. I bet that will fix it then.

Thanks!



"Igor Kuralenok (JetBrains)" <ik@intellij.com> wrote in message
news:bvbai8$er2$1@is.intellij.net...

Grant Gochnauer wrote:

Great. I saw that it was fixed... Thank you so much! :)

>

Its curious why it is having problems finding the class when it is in

the

import, for the class, compiles fine and works on the web app just fine.

Why

would intelliJ be unable to find that class?

Good question :). I think the library containing missing class is not
included in deployment. This condition is relaxed in 1129+. Now TEI
classes is just to be resolved in module scope (in output or libraries).

>

IK

>

--
Igor Kuralenok
Developer
JetBrains Inc.
http://www.jetbrains.com
"Develop with pleasure!"

>


0
Comment actions Permalink

In article <5728279.1075357462441.JavaMail.itn@is.intellij.net>,
Keith Lea <keith@cs.oswego.edu> wrote:

Passwordless SSH key files do work with SourceForge. I think you set it up in
your user preferences, or maybe project preferences.


Now that I have done the proper search at SF, I have found the help
page. Thanks for telling me to look!

(For the curious:
<http://sourceforge.net/docman/display_doc.php?docid=766&group_id=1#sshke
yupload>)

Scott

0
Comment actions Permalink

In article <pan.2004.01.29.03.19.43.226852@talios.com>,
Mark Derricutt <mark@talios.com> wrote:

On Wed, 28 Jan 2004 12:25:26 -0800, Scott Ellsworth wrote:

My biggest problem with the current release is with cvs. It does not
support local access, and the ssh implementation does not handle password
access.


Just run a local pserver. As for SSH and passwords, worked for me just
fine....


Hmm. I had problems with both 1104 and 1120, where any attempt to use
password authentication was met with:

/Users/work/Documents/coding/scott/BonusFinder cvs -n update -d
src/com/iceweasel/bonusFinder/Bonus.java
Command finished ( 6 )

in the cvs output window, and

Error: Valid requests expected

in the status window.

It does work happily, though, if you have an ssh keyfile allowing access
without a passphrase on the cvs server. Now that Keith Lea has pointed
out that Sourceforge allows exactly that kind of access, I am set. I
still would rather it handled passwords for ssh, but it is no longer a
stopship for me.

For what it is worth, I have CVS_RSH set to ssh in Environment.plist,
and the cvs root is :ext:scott@shadow.local:/Users/Shared/coding/cvs/repository

Scott

0

Please sign in to leave a comment.