Ah, maybe. I just saw your "fixes" at my Talk page(s).
xqt
----- Ursprüngliche Nachricht ----- Von: Amir Ladsgroup Gesendet: 26.07.2014 23:22 An: Pywikibot discussion list Betreff: Re: [Pywikipedia-l] The build is broken
it was another bug that John and Ricordisamoa are talking about.
On 7/27/14, info@gno.de info@gno.de wrote:
This is not a bug in testing. The test detects a bug for archivebot.py which fails for unsigned threads.
xqt
----- Ursprüngliche Nachricht ----- Von: John Mark Vandenberg Gesendet: 26.07.2014 18:21 An: Pywikibot discussion list Betreff: Re: [Pywikipedia-l] The build is broken
On Sat, Jul 26, 2014 at 10:39 PM, Ricordisamoa ricordisamoa@openmailbox.org wrote:
The build is still broken, but because of this edit, that breaks the timestamp detection of the ArchiveBot. Maybe a fictitious, write-protected talk page should be used?
That is being tracked on
https://bugzilla.wikimedia.org/show_bug.cgi?id=67663
An alternative test page would be a good solution.
Or maybe the unit test could allow unsigned sections to be added to the talk page.
-- John Vandenberg
Pywikipedia-l mailing list Pywikipedia-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/pywikipedia-l
Sorry for doing that. I had to do it but I didn't have time to check archivebot.py (which is a huge code) and find best fix for it but builds were failing and we were fixing them one by one.
Anyone interested in fixing archivebot.py bug?
Best
On 7/27/14, info@gno.de info@gno.de wrote:
Ah, maybe. I just saw your "fixes" at my Talk page(s).
xqt
----- Ursprüngliche Nachricht ----- Von: Amir Ladsgroup Gesendet: 26.07.2014 23:22 An: Pywikibot discussion list Betreff: Re: [Pywikipedia-l] The build is broken
it was another bug that John and Ricordisamoa are talking about.
On 7/27/14, info@gno.de info@gno.de wrote:
This is not a bug in testing. The test detects a bug for archivebot.py which fails for unsigned threads.
xqt
----- Ursprüngliche Nachricht ----- Von: John Mark Vandenberg Gesendet: 26.07.2014 18:21 An: Pywikibot discussion list Betreff: Re: [Pywikipedia-l] The build is broken
On Sat, Jul 26, 2014 at 10:39 PM, Ricordisamoa ricordisamoa@openmailbox.org wrote:
The build is still broken, but because of this edit, that breaks the timestamp detection of the ArchiveBot. Maybe a fictitious, write-protected talk page should be used?
That is being tracked on
https://bugzilla.wikimedia.org/show_bug.cgi?id=67663
An alternative test page would be a good solution.
Or maybe the unit test could allow unsigned sections to be added to the talk page.
-- John Vandenberg
Pywikipedia-l mailing list Pywikipedia-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/pywikipedia-l
-- Amir
Pywikipedia-l mailing list Pywikipedia-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/pywikipedia-l
From my understanding (I have not tried, just looked at the code),
archivebot.py allows a timestamp to be None. See def should_be_archived(). In that case the thread is not added to the list of threads to be archived but it is left on the talk page. So technically it is not 100% correct to test for a timestamp to be mandatory on the talkpage itself, but only on the threads to be archived.
After all, handling the unsigned case is still TODO in the script.
I think this proposal above by JVB sounds reasonable:
Or maybe the unit test could allow unsigned sections to be added to the talk page.
i.e. to allow also None as test result in line 63 of archivebot_test.py
Mpaa
On Sun, Jul 27, 2014 at 1:44 AM, Amir Ladsgroup ladsgroup@gmail.com wrote:
Sorry for doing that. I had to do it but I didn't have time to check archivebot.py (which is a huge code) and find best fix for it but builds were failing and we were fixing them one by one.
Anyone interested in fixing archivebot.py bug?
Best
On 7/27/14, info@gno.de info@gno.de wrote:
Ah, maybe. I just saw your "fixes" at my Talk page(s).
xqt
----- Ursprüngliche Nachricht ----- Von: Amir Ladsgroup Gesendet: 26.07.2014 23:22 An: Pywikibot discussion list Betreff: Re: [Pywikipedia-l] The build is broken
it was another bug that John and Ricordisamoa are talking about.
On 7/27/14, info@gno.de info@gno.de wrote:
This is not a bug in testing. The test detects a bug for archivebot.py which fails for unsigned threads.
xqt
----- Ursprüngliche Nachricht ----- Von: John Mark Vandenberg Gesendet: 26.07.2014 18:21 An: Pywikibot discussion list Betreff: Re: [Pywikipedia-l] The build is broken
On Sat, Jul 26, 2014 at 10:39 PM, Ricordisamoa ricordisamoa@openmailbox.org wrote:
The build is still broken, but because of this edit, that breaks the timestamp detection of the ArchiveBot. Maybe a fictitious, write-protected talk page should be used?
That is being tracked on
https://bugzilla.wikimedia.org/show_bug.cgi?id=67663
An alternative test page would be a good solution.
Or maybe the unit test could allow unsigned sections to be added to the talk page.
-- John Vandenberg
Pywikipedia-l mailing list Pywikipedia-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/pywikipedia-l
-- Amir
Pywikipedia-l mailing list Pywikipedia-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/pywikipedia-l
-- Amir
Pywikipedia-l mailing list Pywikipedia-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/pywikipedia-l
I disagree because maybe someone breaks the regex of detecting signature date and the test should be able to detect that. The best solution IMHO is to find sections via their revision timestamp but Flow will come and everything will be solved YAYYYY
Best
On 8/2/14, Mpaa mpaa.wiki@gmail.com wrote:
From my understanding (I have not tried, just looked at the code), archivebot.py allows a timestamp to be None. See def should_be_archived(). In that case the thread is not added to the list of threads to be archived but it is left on the talk page. So technically it is not 100% correct to test for a timestamp to be mandatory on the talkpage itself, but only on the threads to be archived.
After all, handling the unsigned case is still TODO in the script.
I think this proposal above by JVB sounds reasonable:
Or maybe the unit test could allow unsigned sections to be added to the talk page.
i.e. to allow also None as test result in line 63 of archivebot_test.py
Mpaa
On Sun, Jul 27, 2014 at 1:44 AM, Amir Ladsgroup ladsgroup@gmail.com wrote:
Sorry for doing that. I had to do it but I didn't have time to check archivebot.py (which is a huge code) and find best fix for it but builds were failing and we were fixing them one by one.
Anyone interested in fixing archivebot.py bug?
Best
On 7/27/14, info@gno.de info@gno.de wrote:
Ah, maybe. I just saw your "fixes" at my Talk page(s).
xqt
----- Ursprüngliche Nachricht ----- Von: Amir Ladsgroup Gesendet: 26.07.2014 23:22 An: Pywikibot discussion list Betreff: Re: [Pywikipedia-l] The build is broken
it was another bug that John and Ricordisamoa are talking about.
On 7/27/14, info@gno.de info@gno.de wrote:
This is not a bug in testing. The test detects a bug for archivebot.py which fails for unsigned threads.
xqt
----- Ursprüngliche Nachricht ----- Von: John Mark Vandenberg Gesendet: 26.07.2014 18:21 An: Pywikibot discussion list Betreff: Re: [Pywikipedia-l] The build is broken
On Sat, Jul 26, 2014 at 10:39 PM, Ricordisamoa ricordisamoa@openmailbox.org wrote:
The build is still broken, but because of this edit, that breaks the timestamp detection of the ArchiveBot. Maybe a fictitious, write-protected talk page should be used?
That is being tracked on
https://bugzilla.wikimedia.org/show_bug.cgi?id=67663
An alternative test page would be a good solution.
Or maybe the unit test could allow unsigned sections to be added to the talk page.
-- John Vandenberg
Pywikipedia-l mailing list Pywikipedia-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/pywikipedia-l
-- Amir
Pywikipedia-l mailing list Pywikipedia-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/pywikipedia-l
-- Amir
Pywikipedia-l mailing list Pywikipedia-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/pywikipedia-l
Recently we had a few cases of code that didnt compile getting merged (I have +2 some of them.. :/), so I have fast tracked the addition of a set of tests which run every script with -help , and with -simulate. These new tests add about 3 minutes to the test suite execution, and add basic validation that the scripts compile and at least main() can be executed.
There are a few scripts which do not emit help on -help, Not too many.
The -simulate argument prevents the scripts from writing to any wiki. Without any other argument, the script should do argument parsing, and usually quit as if called with -help, or provide some informative error message.
Many scripts do not do proper argument parsing and environment sanity checking, resulting in exceptions.
The new tests, annotated with bug numbers, is here: http://git.wikimedia.org/blob/pywikibot%2Fcore.git/master/tests%2Fscript_tes...
The "auto_run_script_list" is the list of scripts which start work without any additional arguments. For those scripts, the tester may wait up to 5 seconds before it kills the process - we may be able to reduce that delay per script by fixing some of the bugs.
Due to some fancy legwork by Legoktm, we now have six Travis builds occurring after each checkin, including running these tests against 1. English Wikipedia, 2. Arabic Wikipedia, and 3. Wikidata.
https://travis-ci.org/wikimedia/pywikibot-core
This means that a month old critical bug is now visible in the two Arabic Wikipedia builds which are failing. There is a patch to be reviewed: https://gerrit.wikimedia.org/r/#/c/149898/
The 'basic' script failing on the py2.6 wikidata build seems to be because py2.6 unit tests are executed in alpha order of the test script name, and the wikidata login doesnt happen earlier for wikidata, but does occur earlier for test scripts against English and Arabic Wikipedia. I am currently working on a fix for this build problem.
-- John Vandenberg
@Amir
If a section is unsigned and there is no timestamp, how can you find sections via their revision timestamp? And expect that the timestamp is a datetime object?
Mpaa
On Sat, Aug 2, 2014 at 8:57 AM, John Mark Vandenberg jayvdb@gmail.com wrote:
Recently we had a few cases of code that didnt compile getting merged (I have +2 some of them.. :/), so I have fast tracked the addition of a set of tests which run every script with -help , and with -simulate. These new tests add about 3 minutes to the test suite execution, and add basic validation that the scripts compile and at least main() can be executed.
There are a few scripts which do not emit help on -help, Not too many.
The -simulate argument prevents the scripts from writing to any wiki. Without any other argument, the script should do argument parsing, and usually quit as if called with -help, or provide some informative error message.
Many scripts do not do proper argument parsing and environment sanity checking, resulting in exceptions.
The new tests, annotated with bug numbers, is here:
http://git.wikimedia.org/blob/pywikibot%2Fcore.git/master/tests%2Fscript_tes...
The "auto_run_script_list" is the list of scripts which start work without any additional arguments. For those scripts, the tester may wait up to 5 seconds before it kills the process - we may be able to reduce that delay per script by fixing some of the bugs.
Due to some fancy legwork by Legoktm, we now have six Travis builds occurring after each checkin, including running these tests against
- English Wikipedia,
- Arabic Wikipedia, and
- Wikidata.
https://travis-ci.org/wikimedia/pywikibot-core
This means that a month old critical bug is now visible in the two Arabic Wikipedia builds which are failing. There is a patch to be reviewed: https://gerrit.wikimedia.org/r/#/c/149898/
The 'basic' script failing on the py2.6 wikidata build seems to be because py2.6 unit tests are executed in alpha order of the test script name, and the wikidata login doesnt happen earlier for wikidata, but does occur earlier for test scripts against English and Arabic Wikipedia. I am currently working on a fix for this build problem.
-- John Vandenberg
Pywikipedia-l mailing list Pywikipedia-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/pywikipedia-l
Mpaa: I meant getting section date by getting history and analyzing revisions instead of checking signature on them.
Best
On 8/2/14, Mpaa mpaa.wiki@gmail.com wrote:
@Amir
If a section is unsigned and there is no timestamp, how can you find sections via their revision timestamp? And expect that the timestamp is a datetime object?
Mpaa
On Sat, Aug 2, 2014 at 8:57 AM, John Mark Vandenberg jayvdb@gmail.com wrote:
Recently we had a few cases of code that didnt compile getting merged (I have +2 some of them.. :/), so I have fast tracked the addition of a set of tests which run every script with -help , and with -simulate. These new tests add about 3 minutes to the test suite execution, and add basic validation that the scripts compile and at least main() can be executed.
There are a few scripts which do not emit help on -help, Not too many.
The -simulate argument prevents the scripts from writing to any wiki. Without any other argument, the script should do argument parsing, and usually quit as if called with -help, or provide some informative error message.
Many scripts do not do proper argument parsing and environment sanity checking, resulting in exceptions.
The new tests, annotated with bug numbers, is here:
http://git.wikimedia.org/blob/pywikibot%2Fcore.git/master/tests%2Fscript_tes...
The "auto_run_script_list" is the list of scripts which start work without any additional arguments. For those scripts, the tester may wait up to 5 seconds before it kills the process - we may be able to reduce that delay per script by fixing some of the bugs.
Due to some fancy legwork by Legoktm, we now have six Travis builds occurring after each checkin, including running these tests against
- English Wikipedia,
- Arabic Wikipedia, and
- Wikidata.
https://travis-ci.org/wikimedia/pywikibot-core
This means that a month old critical bug is now visible in the two Arabic Wikipedia builds which are failing. There is a patch to be reviewed: https://gerrit.wikimedia.org/r/#/c/149898/
The 'basic' script failing on the py2.6 wikidata build seems to be because py2.6 unit tests are executed in alpha order of the test script name, and the wikidata login doesnt happen earlier for wikidata, but does occur earlier for test scripts against English and Arabic Wikipedia. I am currently working on a fix for this build problem.
-- John Vandenberg
Pywikipedia-l mailing list Pywikipedia-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/pywikipedia-l
Analyzing history instead of parsing timestamps from wikitext won't work because it will catch things like {{resolved}} tags that shouldn't delay archiving. It also will keep the common hack of adding a timestamp far in the future to prevent premature archiving from working. On Aug 2, 2014 10:26 AM, "Amir Ladsgroup" ladsgroup@gmail.com wrote:
Mpaa: I meant getting section date by getting history and analyzing revisions instead of checking signature on them.
Best
On 8/2/14, Mpaa mpaa.wiki@gmail.com wrote:
@Amir
If a section is unsigned and there is no timestamp, how can you find sections via their revision timestamp? And expect that the timestamp is a datetime object?
Mpaa
On Sat, Aug 2, 2014 at 8:57 AM, John Mark Vandenberg jayvdb@gmail.com wrote:
Recently we had a few cases of code that didnt compile getting merged (I have +2 some of them.. :/), so I have fast tracked the addition of a set of tests which run every script with -help , and with -simulate. These new tests add about 3 minutes to the test suite execution, and add basic validation that the scripts compile and at least main() can be executed.
There are a few scripts which do not emit help on -help, Not too many.
The -simulate argument prevents the scripts from writing to any wiki. Without any other argument, the script should do argument parsing, and usually quit as if called with -help, or provide some informative error message.
Many scripts do not do proper argument parsing and environment sanity checking, resulting in exceptions.
The new tests, annotated with bug numbers, is here:
http://git.wikimedia.org/blob/pywikibot%2Fcore.git/master/tests%2Fscript_tes...
The "auto_run_script_list" is the list of scripts which start work without any additional arguments. For those scripts, the tester may wait up to 5 seconds before it kills the process - we may be able to reduce that delay per script by fixing some of the bugs.
Due to some fancy legwork by Legoktm, we now have six Travis builds occurring after each checkin, including running these tests against
- English Wikipedia,
- Arabic Wikipedia, and
- Wikidata.
https://travis-ci.org/wikimedia/pywikibot-core
This means that a month old critical bug is now visible in the two Arabic Wikipedia builds which are failing. There is a patch to be reviewed: https://gerrit.wikimedia.org/r/#/c/149898/
The 'basic' script failing on the py2.6 wikidata build seems to be because py2.6 unit tests are executed in alpha order of the test script name, and the wikidata login doesnt happen earlier for wikidata, but does occur earlier for test scripts against English and Arabic Wikipedia. I am currently working on a fix for this build problem.
-- John Vandenberg
Pywikipedia-l mailing list Pywikipedia-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/pywikipedia-l
-- Amir
Pywikipedia-l mailing list Pywikipedia-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/pywikipedia-l
pywikipedia-l@lists.wikimedia.org