qemu-block
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Qemu-block] [Qemu-devel] [PATCH v2 for-2.10 1/1] qemu-iotests: _cle


From: Max Reitz
Subject: Re: [Qemu-block] [Qemu-devel] [PATCH v2 for-2.10 1/1] qemu-iotests: _cleanup_qemu must be called on exit
Date: Wed, 19 Apr 2017 14:45:56 +0200
User-agent: Mozilla/5.0 (X11; Linux x86_64; rv:52.0) Gecko/20100101 Thunderbird/52.0

On 19.04.2017 14:36, Jeff Cody wrote:
> On Wed, Apr 19, 2017 at 02:13:35PM +0200, Max Reitz wrote:
>> On 19.04.2017 12:36, Fam Zheng wrote:
>>> On Tue, 04/18 15:42, Jeff Cody wrote:
>>>>> One idea presented in an earlier thread was that common.rc should create
>>>>> a subdirectory per test (rather than all tests sharing scratch/), and
>>>>> then common.rc itself install the cleanup hook that wipes out the entire
>>>>> subdirectory (or maybe even add a command-line option to ./check to
>>>>> suppress wiping when it is desirable to debug a test failure by seeing
>>>>> the droppings left behind).
>>>>>
>>>>
>>>> Yep, good ideas.  I'll leave that for another series though; there are a 
>>>> few
>>>> other changes based on some suggestions from Daniel that I'd like to
>>>> implement as well, and I can fold your suggestions above into a new series
>>>> (unless someone else does it first).
>>>
>>> Sounds good, I was thinking about making ./check run tests in parallel. If 
>>> each
>>> test has its own tmpdir, it will be much easier.
>>
> 
> Neat - it'd be cool to be able to use the same options as for 'make', e.g.
> ./check -j5.
> 
> 
>> In case you were not yet aware: you can manually set TEST_DIR, this is
>> what I do.
>>
>> (I have a script that basically does TEST_DIR=/tmp/test-$1 ./check -$1)
>>
> 
> Baking it into common without relying on external env variables adds extra
> niceties, because as Eric mentioned cleanup is not left up to the script
> then, and the subdir can be nuked (or kept, via an option to 'check').
> 
> This would make the scripts simpler to read and implement, and less prone to
> do Bad Things. (I keep waiting for some iotest script to misquote a var
> during cleanup and do an rm -f ~/* or something...  it'd be nice to not have
> to worry about that).

Sure, I agree. I just wanted to throw in that it isn't completely
impossible to run iotests in parallel, at least for different formats.

(Baking it in would also mean that we could automatically distribute
test invocations for a single format to multiple instances, yes (as in
your ./check -j5 proposal). But that's work to do. :-))

Max

Attachment: signature.asc
Description: OpenPGP digital signature


reply via email to

[Prev in Thread] Current Thread [Next in Thread]