ignite-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Vyacheslav Daradur <daradu...@gmail.com>
Subject Re: Code inspection
Date Tue, 23 Apr 2019 10:03:46 GMT
>> Also I still strictly against adding checkstyle to project build as minor issues in checkstyle should not be blocker for test run.

Me too.

Looks like everything works fine. Thanks! The only problem is long
time of build ~10 minutes, because of dependencies resolving.

Maxim, please double check it.

On Tue, Apr 23, 2019 at 12:26 PM Petr Ivanov <mr.weider@gmail.com> wrote:
>
> Vyacheslav, can you check this build please [1] if everything was ran as expected?
>
> Also I still strictly against adding checkstyle to project build as minor issues in checkstyle should not be blocker for test run.
>
>
> [1] https://ci.ignite.apache.org/viewLog.html?buildId=3678000&buildTypeId=IgniteTests24Java8_CheckCodeStyle&tab=artifacts&branch_IgniteTests24Java8=%3Cdefault%3E#
>
>
> > On 23 Apr 2019, at 11:59, Petr Ivanov <mr.weider@gmail.com> wrote:
> >
> > I'll check it.
> >
> >
> > Also, please pass TC build for review next time and do not add to Run All without it.
> >
> > Thanks!
> >
> >
> >> On 23 Apr 2019, at 11:53, Vyacheslav Daradur <daradurvs@gmail.com> wrote:
> >>
> >> This is quite strange error, in case of "install" phase it'd be better
> >> just add "checkstyle" profile to "Build" [1] configuration.
> >>
> >> [1] https://ci.ignite.apache.org/viewType.html?buildTypeId=IgniteTests24Java8_BuildApacheIgnite
> >>
> >> On Tue, Apr 23, 2019 at 11:43 AM Petr Ivanov <mr.weider@gmail.com> wrote:
> >>>
> >>> The suite is malformed.
> >>> If no ~/.m2/repository/org/apache/ignite artifact are installed in system, the build will fail [1]
> >>>
> >>> It seems that we should use install instead of validate.
> >>>
> >>>
> >>> [1] https://ci.ignite.apache.org/viewLog.html?tab=buildLog&logTab=tree&filter=debug&expand=all&buildId=3677858&_focus=288
> >>>
> >>>
> >>> On 23 Apr 2019, at 00:25, Vyacheslav Daradur <daradurvs@gmail.com> wrote:
> >>>
> >>> Maxim, I merged your changes to master.
> >>>
> >>> Also, I've created a new build plan "Check Code Style" on TC [1] and
> >>> included in RunAll build.
> >>> The report of check-style attaches in artifacts once build is finished.
> >>>
> >>> Please check that it works as expected once again and announce new
> >>> requirements in a separate thread to avoid confusion of community.
> >>>
> >>> cc Petr, Pavel (JFYI)
> >>>
> >>> [1] https://ci.ignite.apache.org/viewType.html?buildTypeId=IgniteTests24Java8_CheckCodeStyle&tab=buildTypeBranches
> >>>
> >>> On Sun, Apr 21, 2019 at 10:18 PM Vyacheslav Daradur <daradurvs@gmail.com> wrote:
> >>>
> >>>
> >>> Maxim,
> >>>
> >>> I left some comments in Jira, let's resolve them and I'll assist you
> >>> with merge and TC configuring.
> >>>
> >>> On Fri, Apr 19, 2019 at 5:50 PM Maxim Muzafarov <maxmuzaf@gmail.com> wrote:
> >>>
> >>>
> >>> Vyacheslav,
> >>>
> >>> Thank you for your interest in making Ignite coding style better.
> >>>
> >>> The short answer is - there are no different checkstyle
> >>> configurations. One for the JetBrains Inspections, and one the
> >>> Checkstyle plugin. This is a completely different approach for
> >>> checking the Ignites source code.
> >>>
> >>> Currently, we have two different configurations for the JetBrains IDEA
> >>> Inspection check:
> >>> - ignite\.idea\inspectionProfiles\Project_Default.xml - this is
> >>> default on the IDE level and used silently by every developer whos
> >>> checkout Ignite project (it remains)
> >>> - ignite\idea\ignite_inspections_teamcity.xml - this is the
> >>> configuration of the inspection for the TC suite (it will be deleted)
> >>> It's unobvious to maintain both of them. Previously we've planned to
> >>> fix all the inspection rules one by one and add them one by one to the
> >>> TC inspection configuration file (something like storing the
> >>> intermediate result), but it didn't happen cause the inspection TC
> >>> suite got broken after migration to 2018 version.
> >>>
> >>> Now it seems to me, that it is better to use the best open source
> >>> practices like checkstyle plugin (380K usages on github repositories)
> >>> rather than proprietary software. So, we will:
> >>> - keep IDE level inspection configuration (the Project_Default.xml config);
> >>> - add a new checkstyle plugin configuration file (checkstyle.xml
> >>> config) which will be used simultaneously for checking code style on
> >>> build procedure and for the IDE-checkstyle plugin;
> >>>
> >>> On Wed, 17 Apr 2019 at 21:23, Vyacheslav Daradur <daradurvs@gmail.com> wrote:
> >>>
> >>>
> >>> Maxim,
> >>>
> >>> I looked through the PR and it looks good to me in general.
> >>>
> >>> The only question how it's planned to maintain check styles in 2
> >>> different configurations, for IDEA and check style plugin?
> >>>
> >>> On Mon, Mar 25, 2019 at 12:30 PM Maxim Muzafarov <maxmuzaf@gmail.com> wrote:
> >>>
> >>>
> >>> Igniters,
> >>>
> >>> The issue [1] with enabled maven-checkstyle-plugin is ready for the review.
> >>> All changes are prepared according to e-mail [2] the second option
> >>> point (include the plugin in the maven build procedure by default).
> >>>
> >>> JIRA: IGNITE-11277 [1]
> >>> PR: [3]
> >>> Upsource: [4]
> >>>
> >>> How can take a look?
> >>>
> >>> [1] https://issues.apache.org/jira/browse/IGNITE-11277
> >>> [2] http://apache-ignite-developers.2346864.n4.nabble.com/Code-inspection-tp27709p41297.html
> >>> [3] https://github.com/apache/ignite/pull/6119
> >>> [4] https://reviews.ignite.apache.org/ignite/review/IGNT-CR-1018
> >>>
> >>> On Fri, 15 Mar 2019 at 19:19, Dmitriy Pavlov <dpavlov@apache.org> wrote:
> >>>
> >>>
> >>> Hi Igniters,
> >>>
> >>> I see that a new TeamCity is released: Version: 2018.2.3.
> >>>
> >>> Probably it could solve recently introduced problem related to:
> >>> Unexpected error during build messages processing in TeamCity;
> >>>
> >>> Peter I., could you please check?
> >>>
> >>> Sincerely,
> >>> Dmitriy Pavlov
> >>>
> >>> пт, 15 мар. 2019 г. в 12:04, Павлухин Иван <vololo100@gmail.com>:
> >>>
> >>> I agree to gather some votes according to Maxim's proposal.
> >>>
> >>> Personally, I will not put my vote here. Both options will work for me
> >>> today.
> >>>
> >>> But I would like to say a bit about agility. As I said both options
> >>> sounds fine for me today. And I believe that we can switch from one to
> >>> another easily in future. Let's do our best to be flexible.
> >>>
> >>> пт, 15 мар. 2019 г. в 12:04, Павлухин Иван <vololo100@gmail.com>:
> >>>
> >>>
> >>> Maxim,
> >>>
> >>> As far as I understand your case, you want to fix all code styles
> >>>
> >>> issues right before getting the final TC results. Right? ...
> >>>
> >>> Actually, I mostly worry about accidental failures. For simple tasks
> >>> my workflow looks like:
> >>> 1. Create a branch.
> >>> 2. Write some code lines and tests.
> >>> 3. Run the most closely related tests from IDEA.
> >>> 4. Push changes to the branch.
> >>> 5. Launch TC.
> >>> 6. Take a cup of coffee ;-)
> >>> 7. Check TC results after a couple of hours.
> >>>
> >>> And in such workflow I can accidentally leave styling error (IDEA does
> >>> not fail compilation). And I will receive not very valuable report
> >>> from TC. And will have to wait for another couple of hours.
> >>>
> >>> Yes, usually I do not execute "mvn clean install" locally before
> >>> triggering TC. And I think that generally we should not do it because
> >>> TC does it.
> >>>
> >>> If not everybody uses a bot visas it sounds bad for me. For patches
> >>> touching the code it should be mandatory. Also, as you might know
> >>> there are different kind of visas and for some trivial patches we can
> >>> request Checkstyle visa. Committers should check formalities.
> >>>
> >>> пт, 15 мар. 2019 г. в 10:29, Nikolay Izhikov <nizhikov@apache.org>:
> >>>
> >>>
> >>> +1 to enable code style checks in compile time.
> >>>
> >>> We can add option to disable maven codestyle profile with some
> >>>
> >>> environment variable.
> >>>
> >>>
> >>> Anyone who want violate common project rules in their local branch can
> >>>
> >>> set this variable and write some nasty code before push :)
> >>>
> >>>
> >>> пт, 15 марта 2019 г., 9:40 Maxim Muzafarov <maxmuzaf@gmail.com>:
> >>>
> >>>
> >>> Ivan,
> >>>
> >>> 1. I can write code and execute tests successfully even if there are
> >>>
> >>> some style problems. I can imagine that a style error can arise
> >>> occasionally. And instead of getting test results after several hours
> >>> I will get a build failure without any tests run. I will simply lose
> >>> my time. But if the tests are allowed to proceed I will get a red flag
> >>> from code style check, fix those issues and rerun style check.
> >>>
> >>> As far as I understand your case, you want to fix all code styles
> >>> issues right before getting the final TC results. Right? It's doable
> >>> you can disable checkstyle in your local branch and revet it back when
> >>> you've done with all your changes to get the final visa. But the
> >>> common case here is building the project locally and checking all
> >>> requirements for PR right before pushing it to the GitHub repo. I
> >>> always do so. The "Checklist before push" [1] have such
> >>> recommendations. Build the project before push will eliminate your use
> >>> case.
> >>>
> >>> ---
> >>>
> >>> Igniters,
> >>>
> >>> To summarize the options we have with code checking behaviour:
> >>>
> >>> 1)  (code style will be broken more often) Run checkstyle in the
> >>> separate TC suite and include it to the Bot visa.
> >>> - not all of us run TC for their branches especially for simple fixes
> >>> (it's the most common case when a new check style errors occur)
> >>> - not all of us use TC.Bot visa to verify their branches
> >>> - if this checkstyle suite starts to fail it will be ignored for some
> >>> time (not blocks the development process)
> >>> - a lot of suites for code checking (license, checkstyle, something
> >>> else in future)
> >>>
> >>> + a bit comfortable way of TC tests execution for local\prototyped PRs
> >>> (it's a matter of taste)
> >>> + build the project and execute test suites a bit earlier (checkstyle
> >>> on the separate suite does not affect other suites)
> >>>
> >>> 2) (code style will be broken less often) Run checkstyle on project
> >>>
> >>> build stage.
> >>>
> >>> - increases a bit the build time procedure
> >>> - require additional operations to switch it off for prototyping
> >>>
> >>> branches
> >>>
> >>>
> >>> + do not require TC.Bot visa if someone of the community doesn't use
> >>>
> >>> it
> >>>
> >>> + code style errors will be fixed immediately if the master branch
> >>> starts to fail
> >>> + the single place for code checks on maven code validation stage
> >>> (license check suite can be removed)
> >>>
> >>>
> >>> Please, add another advantages\disadvantages that I've missed.
> >>> Let's vote and pick the most suitable option for the Apache Ignite
> >>>
> >>> needs.
> >>>
> >>>
> >>> ---
> >>>
> >>> Personally, I'd like to choose the 2) option.
> >>>
> >>> The JIRA [2] and PR [3] with the checkstyle enabled plugin is ready
> >>> for the review.
> >>>
> >>> [1]
> >>>
> >>> https://cwiki.apache.org/confluence/display/IGNITE/How+to+Contribute#HowtoContribute-Checklistbeforepush
> >>>
> >>> [2] https://issues.apache.org/jira/browse/IGNITE-11277
> >>> [3] https://github.com/apache/ignite/pull/6119
> >>>
> >>> On Thu, 7 Mar 2019 at 11:19, Павлухин Иван <vololo100@gmail.com>
> >>>
> >>> wrote:
> >>>
> >>>
> >>> Maxim,
> >>>
> >>> From use cases described by you I use only 1 and 2. And actually I
> >>> think that we can concentrate on 1 and forget about others for now.
> >>> But please address my worries from previous letter:
> >>> ====Quoted text====
> >>> 1. I can write code and execute tests successfully even if there are
> >>> some style problems. I can imagine that a style error can arise
> >>> occasionally. And instead of getting test results after several
> >>>
> >>> hours
> >>>
> >>> I will get a build failure without any tests run. I will simply lose
> >>> my time. But if the tests are allowed to proceed I will get a red
> >>>
> >>> flag
> >>>
> >>> from code style check, fix those issues and rerun style check.
> >>> 2. Style check takes some time. With simple checks it is almost
> >>> negligible. But it can grow if more checks are involved.
> >>> ====End of quoted text====
> >>>
> >>> Some clarifications. 1 is about running from IDEA first. 2 is
> >>>
> >>> related
> >>>
> >>> to opinions that we should involve much more checks, e.g. using
> >>> abbreviations.
> >>>
> >>> чт, 7 мар. 2019 г. в 10:36, Maxim Muzafarov <maxmuzaf@gmail.com>:
> >>>
> >>>
> >>> Ivan,
> >>>
> >>> Let's take a look at all the options we have (ordered by the
> >>>
> >>> frequency of use):
> >>>
> >>>
> >>> 1. Check ready for merge branches (main case)
> >>> 2. Run tests on TC without checkstyle (prototyping branches)
> >>> 3. Local project build
> >>> 4. Quick build without any additional actions on TC
> >>>
> >>> In the other projects (kafka, netty etc.) which I've checked the
> >>>
> >>> checkstyle plugin is used in the <build> section, so the user has no chance
> >>> in common cases to disable it via command line (correct me if I'm wrong).
> >>> In the PR [1] I've moved checkstyle configuration to the separate profile.
> >>> I've set activation checkstyle profile if -DskipTests specified, so it will
> >>> run with the basic build configuration. It can also be disabled locally if
> >>> we really need it.
> >>>
> >>>
> >>> Back to our use cases:
> >>>
> >>> 1. For checking the ready to merge branches we will fail the
> >>>
> >>> ~Build Apache Ignite~ suite, so no configured checkstyle rules will be
> >>> violated. If we will use the TC.Bot approach someone will merge the branch
> >>> without running TC.Bot on it, but no one will merge the branch with compile
> >>> errors.
> >>>
> >>>
> >>> 2. For the prototyping branches, you can turn off checkstyle in
> >>>
> >>> your local PR by removing activation configuration. It's ok as these type
> >>> of branches will never be merged to the master.
> >>>
> >>>
> >>> 3. From my point, local builds should be always run with the
> >>>
> >>> checkstyle enabled profile. The common build action as `mvn clean install
> >>> -DskipTests` will activate the profile.
> >>>
> >>>
> >>> 4. The checkstyle profile can be disabled explicitly on TC by
> >>>
> >>> specifying -P !checkstyle option. A don't see any use cases of it, but it's
> >>> completely doable.
> >>>
> >>>
> >>> Please, correct me if I've missed something.
> >>>
> >>> I propose to merge PR [1] as it is, with the configured set of
> >>>
> >>> rules.
> >>>
> >>>
> >>> [1] https://github.com/apache/ignite/pull/6119
> >>>
> >>> On Tue, 5 Mar 2019 at 19:02 Павлухин Иван <vololo100@gmail.com>
> >>>
> >>> wrote:
> >>>
> >>>
> >>> Maxim,
> >>>
> >>> I like an idea of being IDE agnostic. I am ok with currently
> >>>
> >>> enabled
> >>>
> >>> checks, they are a must-have in my opinion (even for prototypes).
> >>>
> >>> But I am still do not like an idea of preventing tests execution
> >>>
> >>> if
> >>>
> >>> style check finds a problem. I checked out PR, installed a
> >>>
> >>> plugin and
> >>>
> >>> tried it out. Here are my concerns:
> >>> 1. I can write code and execute tests successfully even if there
> >>>
> >>> are
> >>>
> >>> some style problems. I can imagine that a style error can arise
> >>> occasionally. And instead of getting test results after several
> >>>
> >>> hours
> >>>
> >>> I will get a build failure without any tests run. I will simply
> >>>
> >>> lose
> >>>
> >>> my time. But if the tests are allowed to proceed I will get a
> >>>
> >>> red flag
> >>>
> >>> from code style check, fix those issues and rerun style check.
> >>> 2. Style check takes some time. With simple checks it is almost
> >>> negligible. But it can grow if more checks are involved.
> >>>
> >>> On the bright side I found nice integration of Checkstyle plugin
> >>>
> >>> with
> >>>
> >>> IDEA commit dialog. There is a checkbox "Scan with Checkstyle"
> >>>
> >>> which I
> >>>
> >>> think is quite useful.
> >>>
> >>> пн, 4 мар. 2019 г. в 15:00, Maxim Muzafarov <maxmuzaf@gmail.com
> >>>
> >>> :
> >>>
> >>>
> >>> Ivan,
> >>>
> >>> I like that Jetbrains inspections are integrated with IDE and
> >>>
> >>> TC out
> >>>
> >>> of the box, but currently, they are working not well enough on
> >>>
> >>> TC.
> >>>
> >>> Actually, they are not checking our source code at all.
> >>>
> >>> Let's try a bit another approach and try to be IDE-agnostic
> >>>
> >>> with code
> >>>
> >>> style checking. I've checked popular java projects: hadoop,
> >>>
> >>> kafka,
> >>>
> >>> spark, hive, netty. All of them are using
> >>>
> >>> maven-checkstyle-plugin in
> >>>
> >>> their <build> section by default, so why don't we? It sounds
> >>> reasonable for me at least to try so.
> >>>
> >>> Can you take a look at my changes below?
> >>>
> >>>
> >>> Igniters,
> >>>
> >>> PR [2] has been prepared. All the details I've mentioned in my
> >>>
> >>> comment
> >>>
> >>> in JIRA [4].
> >>> Can anyone take a look at my changes?
> >>>
> >>> JIRA: [1]
> >>> PR: [2]
> >>> Upsource: [3]
> >>>
> >>> Questions to discuss:
> >>> 1) There is no analogue for inspections RedundantSuppression
> >>>
> >>> and
> >>>
> >>> SizeReplaceableByIsEmpty (all code style checks [5]). Propose
> >>>
> >>> to merge
> >>>
> >>> without them.
> >>> 2) Checkstyle plugin has it's own maven profile and enabled by
> >>> default. It can be turned off for prototype branches.
> >>> 3) I've removed the inspections configuration for the TC suite
> >>>
> >>> and
> >>>
> >>> propose to disable it as not working.
> >>>
> >>>
> >>> [1] https://issues.apache.org/jira/browse/IGNITE-11277
> >>> [2] https://github.com/apache/ignite/pull/6119
> >>> [3]
> >>>
> >>> https://reviews.ignite.apache.org/ignite/review/IGNT-CR-1018
> >>>
> >>> [4]
> >>>
> >>> https://issues.apache.org/jira/browse/IGNITE-11277?focusedCommentId=16771200&page=com.atlassian.jira.plugin.system.issuetabpanels%3Acomment-tabpanel#comment-16771200
> >>>
> >>> [5] http://checkstyle.sourceforge.net/checks.html
> >>>
> >>> On Thu, 14 Feb 2019 at 16:21, Павлухин Иван <
> >>>
> >>> vololo100@gmail.com> wrote:
> >>>
> >>>
> >>> Nikolay,
> >>>
> >>> All community members are forced to follow code style.
> >>>
> >>> It's harder to achieve it with dedicated suite.
> >>>
> >>> Why it is easier to follow code style with use of maven
> >>>
> >>> checkstyle
> >>>
> >>> plugin? Is it integrated into IDEA out of box? As I got it
> >>>
> >>> additional
> >>>
> >>> IDEA plugin is needed as well. Who will enforce everybody to
> >>>
> >>> install
> >>>
> >>> it?
> >>>
> >>> Also, as I see a common good practice today is using TC Bot
> >>>
> >>> visa. Visa
> >>>
> >>> includes result from running inspections job.
> >>>
> >>> чт, 14 февр. 2019 г. в 16:08, Nikolay Izhikov <
> >>>
> >>> nizhikov@apache.org>:
> >>>
> >>>
> >>> Ivan,
> >>>
> >>> Could you please outline the benefits you see of failing
> >>>
> >>> compilation and
> >>>
> >>> skipping tests execution if inspections detect a problem?
> >>>
> >>> All community members are forced to follow code style.
> >>> It's harder to achieve it with dedicated suite.
> >>>
> >>>
> >>> чт, 14 февр. 2019 г. в 15:21, Павлухин Иван <
> >>>
> >>> vololo100@gmail.com>:
> >>>
> >>>
> >>> Nikolay,
> >>>
> >>> Should the community spend TC resources for  prototype?
> >>>
> >>> Why not? I think it is not bad idea to run all tests
> >>>
> >>> against some
> >>>
> >>> changes into core classes. If I have a clever idea which
> >>>
> >>> is easy to
> >>>
> >>> test drive I can do couple of prototype-test iterations.
> >>>
> >>> If tests
> >>>
> >>> shows me that everything is bad then the idea was not so
> >>>
> >>> clever and
> >>>
> >>> easy. But if I was lucky then I should discuss the idea
> >>>
> >>> with other
> >>>
> >>> Igniters. I think it is the cheapest way to check the
> >>>
> >>> idea because the
> >>>
> >>> check is fully automated. Requiring a human feedback is
> >>>
> >>> much more
> >>>
> >>> expensive in my opinion.
> >>>
> >>> But, If our code style is not convinient for every day
> >>>
> >>> coding for many
> >>>
> >>> contributors, should you initiate discussion to change
> >>>
> >>> it?
> >>>
> >>> Generally I am fine with our codestyle requirements.
> >>>
> >>> Also, I would like to keep a focus on the subject. Could
> >>>
> >>> you please
> >>>
> >>> outline the benefits you see of failing compilation and
> >>>
> >>> skipping tests
> >>>
> >>> execution if inspections detect a problem?
> >>>
> >>> чт, 14 февр. 2019 г. в 14:14, Nikolay Izhikov <
> >>>
> >>> nizhikov@apache.org>:
> >>>
> >>>
> >>> Hello, Ivan.
> >>>
> >>> Requirements for a prototype code are not the same
> >>>
> >>> as for a patch ready
> >>>
> >>> to merge
> >>>
> >>> True.
> >>>
> >>> I do not see much need in writing good javadocs for
> >>>
> >>> prototype.
> >>>
> >>>
> >>> We, as a community, can't force you to do it.
> >>>
> >>> Why should I stub it to be able run any build on TC?
> >>>
> >>>
> >>> Should the community spend TC resources for  prototype?
> >>> You always can check tests for your prototype locally.
> >>>
> >>> And when it's ready, at least from code style point of
> >>>
> >>> view run it on TC.
> >>>
> >>>
> >>> I, personally, always try to follow project code
> >>>
> >>> style, even for
> >>>
> >>> prototypes.
> >>>
> >>> But, If our code style is not convinient for every day
> >>>
> >>> coding for many
> >>>
> >>> contributors, should you initiate discussion to change
> >>>
> >>> it?
> >>>
> >>>
> >>>
> >>> ср, 13 февр. 2019 г. в 16:45, Павлухин Иван <
> >>>
> >>> vololo100@gmail.com>:
> >>>
> >>>
> >>> Maxim,
> >>>
> >>> Oh, my poor tabs.. Joke.
> >>>
> >>> I am totally ok with currently enabled checks. But I
> >>>
> >>> am mostly
> >>>
> >>> concerned about a general approach. I would like to
> >>>
> >>> outline one thing.
> >>>
> >>> Requirements for a prototype code are not the same
> >>>
> >>> as for a patch
> >>>
> >>> ready to merge (see a little bit more in the end of
> >>>
> >>> that message).
> >>>
> >>>
> >>> We have a document defining code style which every
> >>>
> >>> contributor should
> >>>
> >>> follow [1]. And many points can be checked
> >>>
> >>> automatically. Personally,
> >>>
> >>> I do not see much need in writing good javadocs for
> >>>
> >>> prototype. Why
> >>>
> >>> should I stub it to be able run any build on TC?
> >>>
> >>> Also, we a have a review process which should be
> >>>
> >>> applied to every
> >>>
> >>> patch. Partially it is described in [2]. And due to
> >>>
> >>> this process every
> >>>
> >>> patch should not introduce new failures on TC. So,
> >>>
> >>> the patch should
> >>>
> >>> not be merged if inspections failed.
> >>>
> >>> P.S. Something more about prototypes and production
> >>>
> >>> code. There is a
> >>>
> >>> common bad practice in software engineering. It is
> >>>
> >>> turning prototypes
> >>>
> >>> into production code. Often it is much faster to
> >>>
> >>> create a prototype by
> >>>
> >>> price of violating some rules of writing "clean
> >>>
> >>> code". And often
> >>>
> >>> prototype after successful piloting is turned into
> >>>
> >>> production code.
> >>>
> >>> And it is very easy in practice to keep some pieces
> >>>
> >>> of initially
> >>>
> >>> "dirty" prototype code. I believe human factor plays
> >>>
> >>> a great role
> >>>
> >>> here. How should it be done right then? In my
> >>>
> >>> opinion good production
> >>>
> >>> code should be designed as "good production code"
> >>>
> >>> from the beginning.
> >>>
> >>> So, only ideas are taken from the prototype and a
> >>>
> >>> code is fully
> >>>
> >>> rewritten.
> >>>
> >>> [1]
> >>>
> >>>
> >>> https://cwiki.apache.org/confluence/display/IGNITE/Coding+Guidelines
> >>>
> >>> [2]
> >>>
> >>>
> >>> https://cwiki.apache.org/confluence/display/IGNITE/Review+Checklist
> >>>
> >>>
> >>> ср, 13 февр. 2019 г. в 15:05, Maxim Muzafarov <
> >>>
> >>> maxmuzaf@gmail.com>:
> >>>
> >>>
> >>> Ivan,
> >>>
> >>> As the first implementation of this addition, I'd
> >>>
> >>> prefer to make it
> >>>
> >>> working like _Licenses Headers_ suite check. It
> >>>
> >>> will fail when some
> >>>
> >>> of
> >>>
> >>> the code style checks violated. Moreover, these
> >>>
> >>> licenses header
> >>>
> >>> checks
> >>>
> >>> can be included in the checkstyle plugin
> >>>
> >>> configuration.
> >>>
> >>>
> >>> In general, I'd prefer to have a compilation fail
> >>>
> >>> error with code
> >>>
> >>> style checks and after we will get a stable
> >>>
> >>> checkstyle suite I
> >>>
> >>> propose
> >>>
> >>> to change it in a "compilation error" way. If we
> >>>
> >>> are talking about
> >>>
> >>> the
> >>>
> >>> coding style convenient for most of the community
> >>>
> >>> members I see no
> >>>
> >>> difference with coding sketches or
> >>>
> >>> production-ready branches equally.
> >>>
> >>> Indeed, no one will be against unused imports [or
> >>>
> >>> spaces instead of
> >>>
> >>> tabs :-) ] in their PRs or prototypes, right? (for
> >>>
> >>> instance, it can
> >>>
> >>> be
> >>>
> >>> automatically removed by IDE at commit phase).
> >>>
> >>> Please, note currently enabled checks are:
> >>> - list.isEmpty() instead of list.size() == 0
> >>> - unused imports
> >>> - missing @Override
> >>> - sotred modifiers checks (e.g. pulic static
> >>>
> >>> final ..)
> >>>
> >>> - redundunt suppersion checks
> >>> - spaces insted of tabs.
> >>>
> >>> Are you really what to violate these checks in
> >>>
> >>> your sketches? Hope
> >>>
> >>> not
> >>>
> >>> :-)
> >>>
> >>>
> >>> On Wed, 13 Feb 2019 at 10:25, Nikolay Izhikov <
> >>>
> >>> nizhikov@apache.org>
> >>>
> >>> wrote:
> >>>
> >>>
> >>> Actually, I dont see anything wrong with failing
> >>>
> >>> *compilation*
> >>>
> >>> task.
> >>>
> >>>
> >>> I think one should use project code style for
> >>>
> >>> everyday coding, not
> >>>
> >>> only for
> >>>
> >>> ready-to-merge PRs.
> >>>
> >>> If we cant use code style for everyday coding,
> >>>
> >>> we should change the
> >>>
> >>> codestyle.
> >>>
> >>> What do you think?
> >>>
> >>> ср, 13 февр. 2019 г., 10:11 Petr Ivanov
> >>>
> >>> mr.weider@gmail.com:
> >>>
> >>>
> >>> I guess that was about failing build
> >>>
> >>> configuration with
> >>>
> >>> Checkstype,
> >>>
> >>> not
> >>>
> >>> compilation build itself.
> >>>
> >>> On 12 Feb 2019, at 18:03, Павлухин Иван <
> >>>
> >>> vololo100@gmail.com>
> >>>
> >>> wrote:
> >>>
> >>>
> >>> Folks,
> >>>
> >>> Are you going to fail job compiling Ignite
> >>>
> >>> sources [1] if some
> >>>
> >>> inspection found a problem? Can we avoid it?
> >>>
> >>> It is quite common
> >>>
> >>> pattern to start some feature implementation
> >>>
> >>> with making a
> >>>
> >>> sketch
> >>>
> >>> and
> >>>
> >>> running tests against it. I found it
> >>>
> >>> convenient to skip some
> >>>
> >>> style
> >>>
> >>> requirements for such sketches (e.g. well
> >>>
> >>> formed javadocs).
> >>>
> >>>
> >>> [1]
> >>>
> >>>
> >>>
> >>>
> >>> https://ci.ignite.apache.org/viewType.html?buildTypeId=IgniteTests24Java8_BuildApacheIgnite
> >>>
> >>>
> >>> пн, 11 февр. 2019 г. в 11:38, Nikolay
> >>>
> >>> Izhikov <
> >>>
> >>> nizhikov@apache.org
> >>>
> >>> :
> >>>
> >>>
> >>> Petr, we should have 1 configuration for
> >>>
> >>> project, may be 1
> >>>
> >>> configuration
> >>>
> >>> per programming language.
> >>>
> >>> пн, 11 февр. 2019 г., 11:33 Petr Ivanov
> >>>
> >>> mr.weider@gmail.com:
> >>>
> >>>
> >>> I was asking about how many build
> >>>
> >>> configuration is intended?
> >>>
> >>> One
> >>>
> >>> for
> >>>
> >>> all
> >>>
> >>> and multiple per module?
> >>>
> >>> With IDEA inspections it was going to be
> >>>
> >>> build configuration
> >>>
> >>> per
> >>>
> >>> module.
> >>>
> >>>
> >>>
> >>>
> >>>
> >>> On 11 Feb 2019, at 11:24, Nikolay Izhikov
> >>>
> >>> <
> >>>
> >>> nizhikov@apache.org>
> >>>
> >>> wrote:
> >>>
> >>>
> >>> Hello, Petr.
> >>>
> >>> Are you saying that we have not single
> >>>
> >>> build task? And each
> >>>
> >>> module
> >>>
> >>> builds
> >>>
> >>> when it required? If yes, then I propose
> >>>
> >>> to create a task
> >>>
> >>> like
> >>>
> >>> "Licence
> >>>
> >>> check" which will be run for every patch.
> >>>
> >>> My point is that violation of codestyle
> >>>
> >>> should be treated as
> >>>
> >>> hard as
> >>>
> >>> compile error.
> >>>
> >>> пн, 11 февр. 2019 г., 11:16 Petr Ivanov
> >>>
> >>> mr.weider@gmail.com
> >>>
> >>> :
> >>>
> >>>
> >>> Is build configuration Inspections
> >>>
> >>> [Core] meant to
> >>>
> >>> transform
> >>>
> >>> into
> >>>
> >>> single
> >>>
> >>> all-modules check build configuration
> >>>
> >>> (without module
> >>>
> >>> subdivision)?
> >>>
> >>>
> >>>
> >>> On 11 Feb 2019, at 11:02, Nikolay
> >>>
> >>> Izhikov <
> >>>
> >>> nizhikov@apache.org>
> >>>
> >>> wrote:
> >>>
> >>>
> >>> Hello, Maxim.
> >>>
> >>> +1 from me for migrating to checkstyle.
> >>>
> >>> Oleg, there is plugin for IDEA with
> >>>
> >>> 2mln downloads -
> >>>
> >>>
> >>> https://plugins.jetbrains.com/plugin/1065-checkstyle-idea
> >>>
> >>>
> >>> I propose do the following:
> >>>
> >>> 1. Migrate current checks to checkstyle.
> >>> 2. Apply checks to all Ignite modules.
> >>>
> >>> Currently, only
> >>>
> >>> core
> >>>
> >>> module
> >>>
> >>> are
> >>>
> >>> checked.
> >>> I will review and commit this patch, or
> >>>
> >>> do it by my own.
> >>>
> >>>
> >>> 3. Include code style checks to "Build
> >>>
> >>> Apache Ignite"
> >>>
> >>> suite.
> >>>
> >>> Ignite
> >>>
> >>> has
> >>>
> >>> to
> >>>
> >>> fail to build if patch violates
> >>>
> >>> codestyle.
> >>>
> >>>
> >>> вс, 10 февр. 2019 г. в 07:54, Павлухин
> >>>
> >>> Иван <
> >>>
> >>> vololo100@gmail.com>:
> >>>
> >>>
> >>> Hi,
> >>>
> >>> I also think that some warning from
> >>>
> >>> IDEA that some code
> >>>
> >>> style rule
> >>>
> >>> is
> >>>
> >>> violated is a must-have.
> >>>
> >>> вс, 10 февр. 2019 г. в 01:58,
> >>>
> >>> oignatenko <
> >>>
> >>> oignatenko@gridgain.com
> >>>
> >>> :
> >>>
> >>>
> >>> Hi Maxim,
> >>>
> >>> I believe that whatever style checks
> >>>
> >>> we establish at
> >>>
> >>> Teamcity, we
> >>>
> >>> better
> >>>
> >>> take care of making it easy for
> >>>
> >>> developers to find and
> >>>
> >>> fix
> >>>
> >>> violations
> >>>
> >>> in
> >>>
> >>> their typical dev environment (for
> >>>
> >>> Ignite this means, in
> >>>
> >>> IDEA). I
> >>>
> >>> think
> >>>
> >>> it
> >>>
> >>> is important that developers can
> >>>
> >>> maintain required style
> >>>
> >>> with
> >>>
> >>> minimal
> >>>
> >>> effort
> >>>
> >>> on their side.
> >>>
> >>> If above is doable then I am 200% for
> >>>
> >>> migrating our
> >>>
> >>> Teamcity
> >>>
> >>> inspections
> >>>
> >>> to
> >>>
> >>> checkstyle / maven.
> >>>
> >>> This is because I am very
> >>>
> >>> disappointed observing how it
> >>>
> >>> stays
> >>>
> >>> broken
> >>>
> >>> for
> >>>
> >>> so
> >>>
> >>> long. And worst of all, even when
> >>>
> >>> (if) it is fixed, I
> >>>
> >>> feel
> >>>
> >>> we will
> >>>
> >>> always be
> >>>
> >>> at risk that it breaks again and that
> >>>
> >>> we will have to
> >>>
> >>> again
> >>>
> >>> wait
> >>>
> >>> for
> >>>
> >>> months
> >>>
> >>> for it to be fixed.
> >>>
> >>> This is such a stark contrast with my
> >>>
> >>> experience
> >>>
> >>> regarding
> >>>
> >>> checkstyle
> >>>
> >>> based
> >>>
> >>> inspections. These just work and you
> >>>
> >>> just never fear
> >>>
> >>> that
> >>>
> >>> it is
> >>>
> >>> going
> >>>
> >>> to
> >>>
> >>> break for some obscure reason, this
> >>>
> >>> is so much better
> >>>
> >>> than
> >>>
> >>> what I
> >>>
> >>> observe
> >>>
> >>> now.
> >>>
> >>> One suggestion in case if we pick
> >>>
> >>> checkstyle - I
> >>>
> >>> recommend
> >>>
> >>> keeping
> >>>
> >>> its
> >>>
> >>> config file somewhere in the project
> >>>
> >>> under version
> >>>
> >>> control.
> >>>
> >>> I
> >>>
> >>> used to
> >>>
> >>> maintain such a shared style config
> >>>
> >>> at one of past jobs
> >>>
> >>> and
> >>>
> >>> after
> >>>
> >>> some
> >>>
> >>> experimenting it turned out most
> >>>
> >>> convenient to have it
> >>>
> >>> this
> >>>
> >>> way -
> >>>
> >>> so
> >>>
> >>> that
> >>>
> >>> developers could easily assess and
> >>>
> >>> discuss style
> >>>
> >>> settings
> >>>
> >>> and keep
> >>>
> >>> track
> >>>
> >>> of
> >>>
> >>> changes in these. (note how Kafka
> >>>
> >>> folks from your link
> >>>
> >>> [5]
> >>>
> >>> appear
> >>>
> >>> to
> >>>
> >>> be
> >>>
> >>> doing it this way)
> >>>
> >>> regards, Oleg
> >>>
> >>>
> >>> Mmuzaf wrote
> >>>
> >>> Igniters,
> >>>
> >>> I've found that some of the
> >>>
> >>> community members have
> >>>
> >>> faced
> >>>
> >>> with
> >>>
> >>> `[Inspections] Core suite [1]` is
> >>>
> >>> not working well
> >>>
> >>> enough
> >>>
> >>> on TC.
> >>>
> >>> The
> >>>
> >>> suite has a `FAILED` status for more
> >>>
> >>> than 2 months due
> >>>
> >>> to
> >>>
> >>> some
> >>>
> >>> issues
> >>>
> >>> in TeamCity application [2]. Current
> >>>
> >>> suite behaviour
> >>>
> >>> confuses not
> >>>
> >>> only
> >>>
> >>> new contributors but also other
> >>>
> >>> community members.
> >>>
> >>> Moreover, this
> >>>
> >>> suite is no longer checks rules we
> >>>
> >>> previously
> >>>
> >>> configured.
> >>>
> >>> For
> >>>
> >>> instance, in the master branch, I've
> >>>
> >>> found 11 `Unused
> >>>
> >>> imports`
> >>>
> >>> which
> >>>
> >>> should have been caught earlier
> >>>
> >>> (e.g. for
> >>>
> >>> {{IgniteCachePutAllRestartTest} [3]).
> >>>
> >>> I think we should make the next step
> >>>
> >>> to enable an
> >>>
> >>> automatic code
> >>>
> >>> style
> >>>
> >>> checks. As an example, we can
> >>>
> >>> consider the Apache Kafka
> >>>
> >>> code
> >>>
> >>> style
> >>>
> >>> [5]
> >>>
> >>> way and configure for the Ignite
> >>>
> >>> project a
> >>>
> >>> maven-checkstyle-plugin
> >>>
> >>> with its own maven profile and run
> >>>
> >>> it simultaneously
> >>>
> >>> with
> >>>
> >>> other
> >>>
> >>> TC.
> >>>
> >>> We
> >>>
> >>> can also enable the previously
> >>>
> >>> configured inspection
> >>>
> >>> rules, so no
> >>>
> >>> coding style violations will be
> >>>
> >>> missed.
> >>>
> >>>
> >>> I see some advantages of using a
> >>>
> >>> maven plugin:
> >>>
> >>> - an IDE agnostic way for code checks
> >>> - can be used with different CI and
> >>>
> >>> build tools
> >>>
> >>> (Jenkins,
> >>>
> >>> TC)
> >>>
> >>> - executable from the command line
> >>> - the entry single point to
> >>>
> >>> configure new rules
> >>>
> >>>
> >>> I've created the ticket [4] and will
> >>>
> >>> prepare PR for it.
> >>>
> >>>
> >>> WDYT?
> >>>
> >>> [1]
> >>>
> >>>
> >>>
> >>>
> >>>
> >>>
> >>>
> >>> https://ci.ignite.apache.org/viewType.html?buildTypeId=IgniteTests24Java8_InspectionsCore&branch_IgniteTests24Java8=%3Cdefault%3E&tab=buildTypeStatusDiv
> >>>
> >>> [2]
> >>>
> >>> https://youtrack.jetbrains.com/issue/TW-58504
> >>>
> >>> [3]
> >>>
> >>>
> >>>
> >>>
> >>>
> >>>
> >>>
> >>> https://github.com/apache/ignite/blob/master/modules/core/src/test/java/org/apache/ignite/internal/processors/cache/IgniteCachePutAllRestartTest.java#L29
> >>>
> >>> [4]
> >>>
> >>> https://issues.apache.org/jira/browse/IGNITE-11277
> >>>
> >>> [5]
> >>>
> >>> https://github.com/apache/kafka/tree/trunk/checkstyle
> >>>
> >>>
> >>> On Fri, 21 Dec 2018 at 16:03, Petr
> >>>
> >>> Ivanov &lt;
> >>>
> >>>
> >>> mr.weider@
> >>>
> >>>
> >>> &gt; wrote:
> >>>
> >>>
> >>> It seems there is bug in latest
> >>>
> >>> 2018.2 TeamCity
> >>>
> >>> Bug is filed [1]
> >>>
> >>>
> >>> [1]
> >>>
> >>> https://youtrack.jetbrains.com/issue/TW-58504
> >>>
> >>>
> >>> On 19 Dec 2018, at 11:31, Petr
> >>>
> >>> Ivanov &lt;
> >>>
> >>>
> >>> mr.weider@
> >>>
> >>>
> >>> &gt; wrote:
> >>>
> >>>
> >>> Investigating problem, stand by.
> >>>
> >>>
> >>> On 18 Dec 2018, at 19:41, Dmitriy
> >>>
> >>> Pavlov &lt;
> >>>
> >>>
> >>> dpavlov@
> >>>
> >>>
> >>> &gt; wrote:
> >>>
> >>>
> >>> Both patches were applied. Maxim,
> >>>
> >>> thank you!
> >>>
> >>>
> >>> What about 1. An `Unexpected
> >>>
> >>> error during build
> >>>
> >>> messages
> >>>
> >>> processing in
> >>>
> >>> TeamCity`, what can we do as the
> >>>
> >>> next step to fix
> >>>
> >>> it?
> >>>
> >>>
> >>> Sincerely,
> >>> Dmitriy Pavlov
> >>> [cut]
> >>>
> >>>
> >>>
> >>>
> >>>
> >>>
> >>>
> >>>
> >>> --
> >>> Sent from:
> >>>
> >>>
> >>> http://apache-ignite-developers.2346864.n4.nabble.com/
> >>>
> >>>
> >>>
> >>>
> >>> --
> >>> Best regards,
> >>> Ivan Pavlukhin
> >>>
> >>>
> >>>
> >>>
> >>>
> >>>
> >>>
> >>>
> >>> --
> >>> Best regards,
> >>> Ivan Pavlukhin
> >>>
> >>>
> >>>
> >>>
> >>>
> >>>
> >>> --
> >>> Best regards,
> >>> Ivan Pavlukhin
> >>>
> >>>
> >>>
> >>>
> >>> --
> >>> Best regards,
> >>> Ivan Pavlukhin
> >>>
> >>>
> >>>
> >>>
> >>> --
> >>> Best regards,
> >>> Ivan Pavlukhin
> >>>
> >>>
> >>>
> >>>
> >>> --
> >>> Best regards,
> >>> Ivan Pavlukhin
> >>>
> >>>
> >>> --
> >>> --
> >>> Maxim Muzafarov
> >>>
> >>>
> >>>
> >>>
> >>> --
> >>> Best regards,
> >>> Ivan Pavlukhin
> >>>
> >>>
> >>>
> >>>
> >>> --
> >>> Best regards,
> >>> Ivan Pavlukhin
> >>>
> >>>
> >>>
> >>>
> >>> --
> >>> Best regards,
> >>> Ivan Pavlukhin
> >>>
> >>>
> >>>
> >>>
> >>>
> >>> --
> >>> Best Regards, Vyacheslav D.
> >>>
> >>>
> >>>
> >>>
> >>> --
> >>> Best Regards, Vyacheslav D.
> >>>
> >>>
> >>>
> >>>
> >>> --
> >>> Best Regards, Vyacheslav D.
> >>>
> >>>
> >>
> >>
> >> --
> >> Best Regards, Vyacheslav D.
> >
>


-- 
Best Regards, Vyacheslav D.

Mime
View raw message