Bug 490973 - Broken download pages due to missing "processing"
Summary: Broken download pages due to missing "processing"
Status: RESOLVED FIXED
Alias: None
Product: Platform
Classification: Eclipse Project
Component: Releng (show other bugs)
Version: 4.6   Edit
Hardware: PC Linux
: P3 blocker (vote)
Target Milestone: 4.6 M7   Edit
Assignee: David Williams CLA
QA Contact:
URL:
Whiteboard:
Keywords:
Depends on:
Blocks:
 
Reported: 2016-04-03 23:25 EDT by David Williams CLA
Modified: 2016-04-08 05:03 EDT (History)
2 users (show)

See Also:


Attachments

Note You need to log in before you can comment on or make changes to this bug.
Description David Williams CLA 2016-04-03 23:25:21 EDT
The DL page has missng "BUILD_ID" filled in, and no checksums nor filesize

Platform	                      Download	                Size	Checksum
Windows	                eclipse-SDK-${BUILD_ID}-win32.zip
Windows (x86_64)	eclipse-SDK-${BUILD_ID}-win32-x86_64.zip

While fixing some "testResults Summary" parts of the "GenerateIndex" ant task, I must have removed something that was required.
Comment 1 David Williams CLA 2016-04-03 23:42:14 EDT
I think I have added back a bit too much with this commit, 

http://git.eclipse.org/c/platform/eclipse.platform.releng.buildtools.git/commit/?id=3e9242ee4ef01ce976da9c3916145db88d38dc2f

But, will republish the tools, then try and republish the DL page, to see if any better. 

(A lot of work for something that could easily be done in PHP :)
Comment 2 David Williams CLA 2016-04-04 01:28:30 EDT
It seems I can not "republish" any longer. I have opened bug 490974 for that. 

I will schedule another N-build for around 2 or 3. I'm a bit vague on time since it is possible one of the "running tests" will cause the whole site to be republished. Not sure. 

And, if not that, want to have time to confirm fix on a local build.
Comment 3 David Williams CLA 2016-04-04 02:57:13 EDT
My local build (still) did not publish correctly. 

I've discovered one reason. I thought we no longer used 
org.eclipse.releng.generators.EclipseTestResultsGeneratorNoMail
since neither it nor its super class, 
org.eclipse.releng.generators.EclipseTestResultsGenerator, 
send mail, and the former does not do anything "special". 

So, I ended up making a lot of "protected" or "public" fields and methods private, in the super class, due to thinking we no longer used the subclass. 

I've tried to "restore" most of that, even though much of it is obviously wrong. 
So, we'll see. I'll schedule a build for 3:00 AM, but am not that confident that all is fixed. 

If this latest version is not fixed, I will try to literally remove  more recent versions of "tools" and leave only the "last known working version" in the "build tools" repo (I20160328-0249). (My fear then is there are so many spaghetti interactions between that and the build scripts it may still be broken?). 

The high price of technical debt (and, trying to reduce it!)
Comment 4 David Williams CLA 2016-04-04 11:10:55 EDT
At least the "artifacts" show up again on DL page. 

http://download.eclipse.org/eclipse/downloads/drops4/N20160404-0300/

But the "tests summaries" do not (neither short ones, nor long ones). 
At least that was the part I was trying to fix/improve so is a bit more understandable. 

I'm sure there's a bug for part of the missing "missing files" but, I'll leave this bug open since it's obviously a regression beyond the "small" issue I was trying to fix. .
Comment 5 David Williams CLA 2016-04-04 23:13:57 EDT
Fairly sure this is fixed now. Haven't tested with Equinox or "performance tests", but should be very similar. Will find out soon, as test are about done. :/ 

I did remove the use of the "EclpseTestResultsGeneratorNoMail" class. (neither of them sent mail). But have not yet fixed up all the private/protected/public issues. 

A good illustration of the need for cleanup is this commit that determines which Class is used to "generate indexes". 

http://git.eclipse.org/c/platform/eclipse.platform.releng.aggregator.git/commit/?id=38b310097ae4f3a0f2339643135a98df8bd65e3d

Notice the old code makes frequent mention of EclpseTestResultsGenerator and that class no longer exists. No idea when it was removed, but the main class is 
TestResultsGenerator (no 'Eclipse' at the beginning).
Comment 6 David Williams CLA 2016-04-05 03:00:58 EDT
Still not fixed. 

For the record, DL page (tests area) was still broken initially for 
http://download.eclipse.org/eclipse/downloads/drops4/N20160404-2000/testResults.php

because ".xml" was still on the "configs" set in some ant scripts. 

Fixed that with 
http://git.eclipse.org/c/platform/eclipse.platform.releng.aggregator.git/commit/?id=99724b5fb9a78e28a239164cd6ac773f13e61194

Things were then better, BUT, big problem that even though the tests are put in columns "in order", the order does not match the column labels. 

This can be very confusing. The short summary on first page of DLs is correct, but to know what the "host" really was, one must look in a test result for host. 

By the time we get "all the tests" back, they will be in the right order. But for the case of incomplete results, we need to skip some columns.
Comment 7 David Williams CLA 2016-04-07 09:09:48 EDT
I think we are back to where we were. (Even better, actually).
Comment 8 Markus Keller CLA 2016-04-07 13:20:52 EDT
The new layout without rules makes it hard to follow table rows on http://download.eclipse.org/eclipse/downloads/drops4/N20160406-2000/testResults.php

Changing the <table> attribute rules='groups' to rules='rows' already helps, but the monotonous grid is still harder to use than the structured grouping we had before. Alternating rows colors (with CSS tr:nth-child(even/odd)) could maybe help, but would have to implemented for all tables separately.

This "onHover" rule works with all tables on the testResults:
tr:hover td { background-color: #CCCCFF; }


BTW: I don't think the o.e. prefix is terribly helpful. It's noise, and it still doesn't support find or copy operations. I'd omit it like in the past.
Comment 9 Markus Keller CLA 2016-04-07 13:29:00 EDT
(In reply to Markus Keller from comment #8)
> The new layout without rules makes it hard to follow table rows

https://git.eclipse.org/r/#/c/70161/

This should fix it, but I'm not set up to test it, and I don't know if and how the /staticDropFiles/testResults.php should be updated as well.


> BTW: I don't think the o.e. prefix is terribly helpful. It's noise, and it
> still doesn't support find or copy operations. I'd omit it like in the past.

Fixed with http://git.eclipse.org/c/platform/eclipse.platform.releng.buildtools.git/commit/?id=8fff63dc1b6b2e4c602338ee361592ded25afeff
Comment 10 Eclipse Genie CLA 2016-04-07 14:05:24 EDT
New Gerrit change created: https://git.eclipse.org/r/70161
Comment 12 David Williams CLA 2016-04-07 16:49:55 EDT
Thanks for the tips and fixes, Markus. I do plan to "move to Solstice" sometime which would imply we'd have to use the alternating colors of rows like we do on the downloads page. But in the mean time, "rols" is better. 

Also, since the </table> tag has to be "in the HTML" that is generated, I thought it best to most the <table> tag in their too. 

I also removed some of the no-longer used files and variables. 

I guess I did not explain, that for the "test results" page, we no longer "replace" those %variables% but simply "include" the (pure) HTML files generated 
by the indexer. 

We do still use the "templates" for the drop index.php page, though. 

Changes should be in tonight's nightly.
Comment 13 Dani Megert CLA 2016-04-08 04:08:58 EDT
http://download.eclipse.org/eclipse/downloads/drops4/N20160407-2000/testResults.php

This is very hard to read and there seems to be some problem, see
Releng: Missing testManifest.xml Files
Comment 14 Dani Megert CLA 2016-04-08 04:19:58 EDT
> This "onHover" rule works with all tables on the testResults:
> tr:hover td { background-color: #CCCCFF; }

Helps, but not when simply looking at the table.
Comment 15 David Williams CLA 2016-04-08 05:03:21 EDT
Since this bug was a "blocker", originally about the pages being *broken*, I think we should handle other issues in specific bugs. I have reopened 491117 since my improvements there did not work as expected. 

I have newly opened bug 491304 to handle any other improvements to "look and feel" of the page. (And, I am sure there are many open on that topic already).