-
Notifications
You must be signed in to change notification settings - Fork 33
logbook
TODO: https://github.com/php-coder/mystamps/issues/1267#issuecomment-594005561
TODO: add 8ec711fcf1fc2918262bbfa821d56ccaa9767b65 (emulating <fmt:formatNumber>)
TODO: add https://github.com/php-coder/mystamps/issues/246 (bootstrap-languages)
TODO: mention truncated logs
TODO: mention transition to trusty and broken MySQL
TODO: CSP and unsafe-inline
TODO: CSP and unsafe-eval
TODO: mention about dependency between mocks and /image/1
TODO: https://github.com/yegor256/0pdd/issues/206 and eaa432546975f44e23b09c21f9740161fb90092d
19.04.2019
During upgrade of H2 to a newer version, one of the Liquibase migrations has started to fail because of precondition hasn't been met. It seems like a bug in H2 1.4.198 and 1.4.199 I've rolled back to the working version (1.4.197) and submitted the issue to upstream.
Issue: https://github.com/h2database/h2database/issues/1874
UPDATE(22.04.2019): after investigation it has been found that it was caused by the SQL query with undefined behavior (that has been changed in the latest H2 versions). I've fixed the query in cdeec7fb4
28.02.2019
During upgrade of htmlunit to a newer version, some tests have started to fail because uploaded image (test.png) now has an incorrect application/octet-stream content type. It happens because all emulated browsers don't support this image type. I've applied a workaround by renaming a file to test.jpg because JPEG format isn't affected by this bug.
Bug: https://github.com/HtmlUnit/htmlunit/issues/18
TODO: add commit with workaround (when it will be merged to master)
TODO: revert workaround when bug will be fixed
16.01.2019
I tried to use errorprone-slf4j plugin but it fails with an error and without any clue on what is wrong. I've reported a bug.
Issue: https://github.com/KengoTODA/errorprone-slf4j/issues/22
TODO: start using this plugin when it will be possible to use it
12.01.2019
A newer version of 0pdd brought a new version of pdd that has a backward incompatible change: it has started to treat simple TODO comments as pdd instructions. Because we have many of them across the project and none of them are intented to satisfy pdd requirements, the latter started to fail a build. I've replaced TODO work in comments by FIXME to make it work again.
Yet another regression has been identified: pdd started to complain on XML attribute like value="TODO|XXX|NOPMD". I've filled a bug about that.
Issue: https://github.com/yegor256/0pdd/issues/260
Issue: https://github.com/yegor256/pdd/issues/117
TODO: revert a9ebbc44d (and probably ef7b41ee4)
TODO: revert ff2556312 when pdd #117 will be fixed
Commits: ef7b41ee4, ff2556312, a9ebbc44d
06.10.2018
During migration to Selenium3, I faced the following error:
java.lang.NoSuchMethodError: org.apache.commons.io.IOUtils.toString(Ljava/io/InputStream;Ljava/nio/charset/Charset;)Ljava/lang/String;
It has happened because of a conflict between dependencies: a newer htmlunit uses a method from commons-io that was introduced in version 2.3. At the same time, robotframework-maven-plugin depends on plexus-io:2.7 that pulls commons-io:2.2 Because the old version of the library was in a classpath, the tests were failing with that error. In order to fix this, I made a fork of robotframework-maven-plugin, that has updated version of plexus-io (and commons-io as a consequence). Until my pull request isn't merged, we will use a forked version of the plugin.
TODO: switch back to original version of robotframework-maven-plugin
Pull Request: https://github.com/robotframework/MavenPlugin/pull/47
Commits: TODO
06.10.2018
During integration of the dockerfile-maven-plugin, I faced up a limitation that I can't have Dockerfile in a separate directory. I wanted to continue to use docker directory but the plugin wants to have it in the root directory or at least in the same directory as a WAR file. There is a related pull request that adds support for a separate directory but it's not merged yet. Until it's merged, we will use a workaround: mvn package now also copies docker/Dockerfile into target and that directory is being used as a docker context.
TODO: remove this workaround
Pull Request: https://github.com/spotify/dockerfile-maven/pull/89
Commits: c09aa64814
14.12.2017
0pdd complained on @todo marker inside a comment in the checkstyle.xml file. I filled an issue and got an advice to replace it by @todo. I applied this workaround.
Commit: c6899e756
Issue: https://github.com/yegor256/0pdd/issues/168
22.09.2017
After enabling Content Security Policy, I've started to get errors from a browser on the H2 web console about using style attribute. I added workaround for them by using hash-sums and also submitted an issue to the upstream.
Commits: bd7216d29
Issue: https://github.com/h2database/h2database/issues/612
21.09.2017
After enabling Content Security Policy, I've started to get errors from a browser on the Togglz web console about using style attribute. I added workaround for them by using hash-sums and also submitted an issue to the upstream.
Commits: d9b322ecf
Issue: https://github.com/togglz/togglz/issues/248
UPDATE(15.12.2017): one of the style attributes gone after I updated Togglz to 2.5.0 version in the commit 47a2dffd0
UPDATE(19.08.2018): the last style attribute was removed in Togglz 2.6.0 version in the commit 9c896a6d1
07.09.2017
TravisCI environment has been changed and now it contains _JAVA_OPTIONS="-Xmx2048m -Xms512m" variable. This innocent change breaks our build because html5validator shows warning about that variable. I added a workaround for that and reported to the upstream.
Commits: 69cb4df5d
Issues: https://github.com/php-coder/mystamps/issues/634
Issue: https://github.com/svenkreiss/html5validator/issues/34
UPDATE(14.01.2018): workaround has been removed in the commit 242d5ed37 as this warning were fixed by upstream in the 0.2.8 version.
26.08.2017
CodeNarc produces SpaceAfterComma false positive. I've reported about it to the upstream.
TODO: remove these suppressions when the issue will be fixed in the upstream
Commits: 3878ba2d9
Issue: https://github.com/CodeNarc/CodeNarc/issues/225
01.08.2017
Buggy TravisCI has started to trim the build logs. To be able to find out why Robot Framework tests fail, I tried to reduce output from the builds by gzipping the report. Later I also decided to use XML report that contains just a raw data. Unfortunately, it didn't help but the changes are good to having them permanently.
Issue: https://github.com/travis-ci/travis-ci/issues/8189
13.07.2017
It turned out that Thymeleaf Tooglz dialect has a regression that prevents it from removing xmlns:togglz attribute from the document. It has been reported to the upstream.
Issue: https://github.com/heneke/thymeleaf-extras-togglz/issues/10
UPDATE(23.08.2017): PR has been submitted: https://github.com/heneke/thymeleaf-extras-togglz/pull/11
UPDATE(23.08.2017): workaround has been added: bbc3f66f5
UPDATE(27.08.2017): a new version was released and workaround has been removed in commit 2d98048dd
30.05.2017
When authenticated user is opening //series/add URL, an exception org.thymeleaf.exceptions.TemplateInputException: Error resolving template "/series/add", template might not exist or might not be accessible by any of the configured Template Resolvers occurs. Because this looks like a but in the Spring Framework, I've created an issue for that.
Issue: https://github.com/php-coder/mystamps/issues/599
Issue: https://jira.spring.io/browse/SPR-15596
27.05.2017
During writing a test, it turned out that the keyword Textfield Value Should Be from Selenium2Library fails with NullPointerException, when it's used for getting a value from the <input type=url>. I've filled an issue to the upstream.
TODO: remove this workaround when it will be fixed in the upstream
Issue: https://github.com/MarkusBernhardt/robotframework-selenium2library-java/issues/92
Issue: https://github.com/Hi-Fi/robotframework-seleniumlibrary-java/issues/52
UPDATE(17.10.2017): workaround has been added.
UPDATE(13.08.2018): workaround has been extracted into separate keyword in the commit 03d6175a3
UPDATE(20.08.2018): after migrating to Hi-Fi/robotframework-seleniumlibrary-java library, the problem is still here. I have created an issue for that.
15.04.2017
html5validator has started to fail with This document appears to be written in Danish but the "html" start tag has "lang="en" error since 0.27 version. Workaround has been added by ignoring such warnings.
TODO: remove this workaround when false positive will be fixed in the upstream
Commits: a09a3f523, 0abdb8103, 27d56377b
Issue: https://github.com/svenkreiss/html5validator/issues/32
UPDATE(29.12.2017): it still there and I created an issue in our tracker: https://github.com/php-coder/mystamps/issues/783
UPDATE(14.01.2018): in the commit 883918435, I removed all workarounds and has started to use a new --no-langdetect option that was introduced in html5validator 0.2.10
06.04.2017
Thymeleaf Tooglz dialect doesn't have an easy way to set an attribute value depends on the feature state. I had to use 2 tags with different togglz:active/togglz:inactive attributes.
UPDATE(19.04.2017): I decided to not use Togglz for an attribute but move it to the controller. Approach with 2 tags didn't work because integration tests were failing (they were expecting a new URL but server was returning an old one).
Issue: https://github.com/php-coder/mystamps/issues/387
Issue: https://github.com/heneke/thymeleaf-extras-togglz/issues/9
05.04.2017
danger has started to fail with 'find_merge_base': Cannot find a merge base between danger_base and danger_head on pull requests. To workaround this error I rolled back to last working version 4.3.1
Issue: https://github.com/danger/danger/issues/768
UPDATE(21.07.2017): I suspect that it caused by the 08bb9a48b commit where I decreased the depth of git clone. In order to validate this theory I've reverted all 3 commits (see a922225bc, b7f42d326, a502b7e17).
14.03.2017
John Shkarin has reported that checkstyle on Windows complains about a lot of errors. We suspect that this happens because src/main/config/checkstyle-suppressions.xml file isn't taking into account at all or some of the suppressions are ignored (probably because of / vs \). We tried to follow the instructions from this blog but it didn't help.
Report: https://github.com/php-coder/mystamps/pull/547#issuecomment-286550307
UPDATE(04.08.2018): fixed in e15abf3e8
01.03.2017
Input Text keyword from Selenium2Library doesn't support Russian characters in the report. I've created a bug report.
Issue: https://github.com/MarkusBernhardt/robotframework-selenium2library-java/issues/90
UPDATED(23.09.2019): I've re-submitted it to the new upstream: https://github.com/Hi-Fi/robotframework-seleniumlibrary-java/issues/92
27.02.2017
Selenium2Library shows poor error message when we're trying to interact with an invisible element: You may only interact with visible elements. It doesn't contain locator of this element, for example. I've created an issue for that in upstream but author said that it should be implemented in the Selenium itself.
Issue: https://github.com/robotframework/SeleniumLibrary/issues/769
UPDATED(26.04.2017): I've created this in the Selenium: https://github.com/SeleniumHQ/selenium/issues/3917
26.01.2017
When Selenium2Library and htmlunit are used together, the latter prints a lot of messages to the console. I asked upstream but didn't get an answer. To workaround this, I created a script that is filtering out these noise messages from the output.
Issue: https://github.com/MarkusBernhardt/robotframework-selenium2library-java/issues/85
Issue: https://github.com/php-coder/mystamps/issues/538
UPDATE(30.04.2019): the issues has been filled against RobotMaven plugin: https://github.com/robotframework/MavenPlugin/issues/65
UPDATE(21.01.2020): after debugging this deeper it has been turned out that the plugin uses logger configuration from Maven that in turn use slf4j and its SimpleLogger. It has been configured to not output these annoying messages and the workaround has been removed.
Commits: e6ed0e855
05.10.2016
Because codenarc-maven-plugin doesn't pick up codenarc.properties file, I had to use @SuppressWarnings for ignoring some rules.
TODO: these workarounds should be removed when we'll find a way of using codenarc.properties
Commits: 24c47c2d6
Issue: https://github.com/gleclaire/codenarc-maven-plugin/issues/14
Issue: https://github.com/php-coder/mystamps/issues/539
05.10.2016
CodeNarc produces false positive on Spock tests that uses || as a delimiter for test data. I've reported about it to the upstream.
TODO: remove these workarounds when the issue will be fixed in the upstream
Commits: 24c47c2d6
Issue: https://github.com/CodeNarc/CodeNarc/issues/176
30.09.2016
Deploy to prod with ansible has started to fail with {"assertion": "war_file.stat.exists", "changed": false, "evaluated_to": false, "failed": true} To workaround this error I rolled back to last working version 2.1.1.0
Commits: fc72d3b93
Issue: https://github.com/ansible/ansible-modules-core/issues/5115
UPDATE(21.09.2017): the new address of the issue is https://github.com/ansible/ansible/issues/29618
UPDATE(10.11.2019): this problem has been resolved in the 1bf3966bf commit by using the paths that are relative to the playbook instead of relying on the current directory.
19.01.2016
When I tried to switch to using docker based infrastructure in Travis CI, integration tests has started to fail because Spring application did not start before the configured timeout (30000ms). Increasing a log levels didn't help to reveal root cause. I've asked Spring Boot developers on chat and they said that they also tried to use docker infrastructure in Travis and gave up at the end. So I decided to stay with the non-docker based version too.
Build log: https://travis-ci.org/php-coder/mystamps/jobs/103406690
Build log with enabled debug messages: https://travis-ci.org/php-coder/mystamps/jobs/103409336
04.02.2014
MultipartFile.transferTo() from Spring Framework works incorrectly with Jetty: it saves a file relative to a directory from multipart-config/location (in web.xml). As a result instead of saving file to /data/uploads it tries to write it to /tmp/data/uploads. I added a workaround for this behavior by not using transferTo() method and writing a file manually.
TODO: remove this workaround because it was fixed in Spring Framework 4.3.8 (see https://jira.spring.io/browse/SPR-15257)
Commit: a3bb652ff