tag:help.appveyor.com,2012-11-13:/discussions/problems/3510-same-commit-builds-fail-on-one-acc-but-not-the-otherAppVeyor: Discussion 2018-08-25T02:01:58Ztag:help.appveyor.com,2012-11-13:Comment/384087732015-11-09T21:33:21Z2015-11-09T21:33:21Zsame commit, builds fail on one acc but not the other<div><p><code>HaxeFoundation</code> account is running on new
environment (with greater and more predictable performance) where
build worker images were re-installed from scratch.
<code>andyli</code> is running on current Hyper-V based environment
which will be finally migrated to a new one in 1.5 weeks.</p>
<p>Looks like something is missing/different on that new build
workers. What does that test do? Can it be simplified to few lines?
Is it language-specific. Can you please help us to figure out
that?</p></div>Feodor Fitsnertag:help.appveyor.com,2012-11-13:Comment/384087732015-11-09T21:49:26Z2015-11-09T21:49:26Zsame commit, builds fail on one acc but not the other<div><p>Interesting. I've just run your project on my test account and
it worked: <a href="https://ci.appveyor.com/project/FeodorFitsner/haxe">https://ci.appveyor.com/project/FeodorFitsner/haxe</a></p>
<p>Environment is the same, but the only difference is VM
configuration. Mine was 2 cores with 7 GB and yours - 1 core with
3.5 GB. Looks like it makes a difference.</p>
<p>I've changed a plan for <code>HaxeFoundation</code> account to
run on 2 cores. Let me know how it worked for you.</p></div>Feodor Fitsnertag:help.appveyor.com,2012-11-13:Comment/384087732015-11-09T21:51:07Z2015-11-09T21:51:07Zsame commit, builds fail on one acc but not the other<div><p>In case it worked for you on 2 cores... I have a follow up
question - why? Does that require at least 2 cores?</p></div>Feodor Fitsnertag:help.appveyor.com,2012-11-13:Comment/384087732015-11-10T08:26:50Z2015-11-10T08:26:50Zsame commit, builds fail on one acc but not the other<div><p>Umm.. The builds in HaxeFoundation account now work as
expected.<br>
I don't think the test requires at least 2 cores - it fails when
trying to write a new file, and it was a permission error.</p>
<p>I've saved the python output of the haxe compiler, reduced it to
run only the failing case, added some more debug message, and
pushed it to <a href="https://github.com/andyli/appveyor-debug">https://github.com/andyli/appveyor-debug</a><br>
But I don't have the failing appveyor environment (1 core with 3.5
GB), so I cannot test whether it will fail.<br>
Here is the build using my andyli account: <a href="https://ci.appveyor.com/project/andyli/appveyor-debug/build/1.0.4">
https://ci.appveyor.com/project/andyli/appveyor-debug/build/1.0.4</a></p></div>andytag:help.appveyor.com,2012-11-13:Comment/384087732015-11-10T18:32:24Z2015-11-10T18:32:24Zsame commit, builds fail on one acc but not the other<div><p>Here you are - account <code>andyli</code> has been moved to the
plan with 1 core and 3.5 GB of ram. Let me know about your results
please.</p></div>Feodor Fitsnertag:help.appveyor.com,2012-11-13:Comment/384087732015-11-12T10:06:25Z2015-11-12T10:06:25Zsame commit, builds fail on one acc but not the other<div><p>I've investigated and it seems it is my fault. The problem is,
there is a loop in which I write, execute, and delete a file. The
error happens during the second time I write the file with the same
name as the previously deleted one. If I add a
<code>time.wait(0.5)</code> between the first delete and the second
write, it works as expected. So it seems it is just there is some
time between I delete a file and the file is actually removed from
the system.</p></div>andytag:help.appveyor.com,2012-11-13:Comment/384087732015-11-12T17:41:07Z2015-11-12T17:41:07Zsame commit, builds fail on one acc but not the other<div><p>Nice finding, thanks for looking in that!</p>
<p>I'm just wondering regarding the overall build time on 1 core vs
2 cores - do you see any performance gain on 2 cores?</p></div>Feodor Fitsnertag:help.appveyor.com,2012-11-13:Comment/384087732015-11-12T18:32:42Z2015-11-12T18:32:42Zsame commit, builds fail on one acc but not the other<div><p>Yes, there is a drop in build time from ~16 min to ~13min. I
guess it could be even faster if we enable parallel build for the
C++ compilation stage.</p></div>andytag:help.appveyor.com,2012-11-13:Comment/384087732015-11-12T18:33:52Z2015-11-12T18:33:52Zsame commit, builds fail on one acc but not the other<div><p>OK, nice, good to know.</p></div>Feodor Fitsner