SA Bugzilla – Bug 857
Don't do razor tests by default
Last modified: 2002-10-18 01:43:46 UTC
I'm going to disable the razor tests by default. They are just down *way* too often.
Subject: Re: [SAdev] New: Don't do razor tests by default OK, now these are disabled by default, but prompts the user if he/she wants to run the razor tests (for both razor1 and razor2). I'm not enamoured by prompting Makefile.PL's, so alternate suggestions are welcome.
I'd rather fix our code to deal more gracefully with server failures and I would like the razor tests to remain on by default. As far as I am concerned, the bug is in our code.
Subject: Re: [SAdev] Don't do razor tests by default On Mon, Sep 09, 2002 at 01:10:29PM -0700, bugzilla-daemon@hughes-family.org wrote: > I'd rather fix our code to deal more gracefully with server failures and > I would like the razor tests to remain on by default. > > As far as I am concerned, the bug is in our code. Not really, the bug is generated (typically, for the "bad file descriptor" error) by IO::Select, which causes the Razor2::Client code to die, so we catch it, display the error, and continue. It's not a bug in our code unless you can interpret the death string and want to print something else out. You won't be able to fix the error unless you dig into IO::Select and/or Razor2::Agent modules... There's a reason I had to submit Razor2 patches while trying to make R2 work with SA.
Subject: Re: [SAdev] Don't do razor tests by default felicity@kluge.net writes: > Not really, the bug is generated (typically, for the "bad file > descriptor" error) by IO::Select, which causes the Razor2::Client > code to die, so we catch it, display the error, and continue. Okay, but it is still our problem, not Razor's. If it's a bug in Perl, then we have to deal with it, work around it, fix it, etc. We shouldn't be relying on servers to be 100% reliable; I don't want us to turn off the razor tests by default.
Subject: Re: Don't do razor tests by default > Okay, but it is still our problem, not Razor's. If it's a bug in > Perl, then we have to deal with it, work around it, fix it, etc. We > shouldn't be relying on servers to be 100% reliable; I don't want us > to turn off the razor tests by default. I was trying to follow the model that other perl network modules take - offer network tests as an option, but don't perform them by default. See for example LWP, etc. At the moment we fail the test if Razor is down, and this is absolutely terrible given how unstable Razor is. And you don't get the CPAN fail reports - I do ;-) Matt.
I'm closing this bug. it's actually worse than Matt sez. here's why: our Razor tests test to see if a spam is reported in the razor servers. 1. the spam may not have been nominated recently, or 2. the user may not have a high enough "reporter rating" to nominate it, or 3. the report/catalogue servers may be out of sync, which they frequently are, or 4. the user may not have registered before running the test, or 5. the servers may be down. All these would cause the test to return a "fail". This means that our Razor tests are unreliable, and should be off by default. The script just cannot test Razor effectively. I'm not saying let's not support Razor, just let's not include it in the default set of tests that users get when they install SA. so I'm going to make it off by default in 2.43, and leave it as a prompt in 2.50 (overrideable by cmdline for Duncan).
Actually, here's a proposal. The prompting is nasty, and I'd like to make it go away as well. So how's about we remove Razor tests from the default "make test", and make a new target for the Makefile, "testall", which runs "make test" and the Razor tests? sound better?
*** Bug 1006 has been marked as a duplicate of this bug. ***
Subject: Re: Don't do razor tests by default > Actually, here's a proposal. The prompting is nasty, and I'd like > to make it go away as well. So how's about we remove Razor tests > from the default "make test", and make a new target > for the Makefile, "testall", which runs "make test" and the Razor > tests? > > sound better? Well the problem with that is nobody will ever run it ;-) I hate prompting too, but it seems the de-facto way to do it (LWP/LibNET for example).
We shouldn't turn off testing for anything we ship in SA. Otherwise, it will break. If the problem is the server being down, then change the test to be ignored if the server is verifiably down.
Subject: Re: [SAdev] Don't do razor tests by default the server being down is only 1/2 the problem. as far as I can see, most of the time nowadays the problem is that the reporting server (which you report to) does not sync with the catalogue server (which you then check against), so the test will fail due to that; unless you sleep for whatever the syncing period is, which could be about 15 minutes, and depends entirely on the Razor backend. --j.
Subject: Re: [SAdev] Don't do razor tests by default On Tue, Oct 15, 2002 at 03:20:08AM -0700, bugzilla-daemon@hughes-family.org wrote: > from the default "make test", and make a new target > for the Makefile, "testall", which runs "make test" and the Razor > tests? Do we do tests for things like dcc and pyzor as well? I don't understand why we'd want to test Razor beyond making sure we could load the module. Anything past that is really not our issue.
we probably shouldn't bother testing anything apart from "the module loads". it can be hard to tell though. anyway, this bug is fixed ;)