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

Re: Checking build errors <- endianess problem



Dear Gert,

The output you pasted in your previous e-mail shows that the checks fail
when comparing output based on ASCII input files with output based on
the binary input files.

On 23-01-14 15:20, Gert Wollny wrote:
> Hello, 
> 
> I dug around in the code and voila, e.g. in fvlib/frutil.cpp the
> function blockWriteOrRead uses fstream.read|write to do raw data IO and
> then in other parts of the code the data is just cast to the desired
> type without doing any checks of endianess let alone the needed
> conversions.

Endianness! I should have thought of that. I think the underlying
library was never developed with endianness in mind. Maybe I can
convince upstream to fix this. Not sure how useful that would be as I
don't expect many users running the kind of analysis ProbABEL is
designed for on a different architecture...

> 
> Now one can either fix the code, or mask the big-endian archs in the
> control file.

That will at least be the short-term solution.


Thanks for you time and effort,

Lennart.

> 
> regards, 
> Gert 
> 
> 
> On Thu, 2014-01-23 at 12:39 +0100, Gert Wollny wrote:
>> Hi, 
>>
>> I have access to a powerpc laptop and took the liberty to run the build.
>> See the testlog below.
>>
>> Considering that some of the inputfiles seem to be binary, I would guess
>> that this problem stems from powerpc (and mips?) being big-endian. 
>>
>> hope that helps, 
>> Gert 
>>
>>
>>
>> ===========================================
>>    ProbABEL 0.4.2: checks/test-suite.log
>> ===========================================
>>
>> # TOTAL: 6
>> # PASS:  2
>> # SKIP:  0
>> # XFAIL: 0
>> # FAIL:  4
>> # XPASS: 0
>> # ERROR: 0
>>
>> .. contents:: :depth: 2
>>
>> FAIL: test_qt.sh
>> ================
>>
>> Analysing QT...
>> base analysis
>> file contains data of unknown type 1280
>> diff: linear_base_fv_add.out.txt: Datei oder Verzeichnis nicht gefunden
>> QT check: dose vs. dose_fv
>> FAILED
>>
>> FAIL: test_mms.sh
>> =================
>>
>> Analysis using MMScore...
>> file contains data of unknown type 1280
>> diff: mmscore_dose_fv_add.out.txt: Datei oder Verzeichnis nicht gefunden
>> mmscore check: dose vs. dose_fv
>> FAILED
>>
>> FAIL: test_bt.sh
>> ================
>>
>> Analysing BT...
>> file contains data of unknown type 1280
>> diff: logist_fv_add.out.txt: Datei oder Verzeichnis nicht gefunden
>> BT check: dose vs. dose_fv
>> FAILED
>>
>> FAIL: test_cox.sh
>> =================
>>
>> Analysing Cox model...
>> Warning for rs3174230: beta may be infinite, setting beta and se to
>> 'nan'
>> file contains data of unknown type 1280
>> diff: coxph_dose_fv_add.out.txt: Datei oder Verzeichnis nicht gefunden
>> pacoxph check: dose vs. dose_fv
>> FAILED
>>
>>
>>
>> On Thu, 2014-01-23 at 07:27 +0800, Paul Wise wrote:
>>> On Thu, Jan 23, 2014 at 3:59 AM, L.C. Karssen wrote:
>>>
>>>> What I can't see is their exact error messages. This is because automake
>>>> by default runs tests in parallel and sends the output of the individual
>>>> tests to a file called test-suite.log. Is there a way to get access to
>>>> that file on the build machine?
>>>> Or is there another way to find out why the checks fail (short of
>>>> uploading a package where I run the tests serially and send output to
>>>> the screen)?
>>>
>>> You can choose one in decreasing order of preference:
>>>
>>> Ask your sponsor to login to the porterboxen, do a build and send you
>>> the results.
>>>
>>> https://db.debian.org/machines.cgi
>>>
>>> Get access to the porterboxen yourself, do a build and inspect the results.
>>>
>>> https://dsa.debian.org/doc/guest-account/
>>>
>>> Override dh_auto_test to output test-suite.log when it fails, reupload.
>>>
>>>> Please CC me in any replies as I'm not subscribed to this list.
>>>
>>> Done (I'm subscribed).
>>>
>>> -- 
>>> bye,
>>> pabs
>>>
>>> https://wiki.debian.org/PaulWise
>>>
>>>
>>
>>
> 
> 

-- 
-----------------------------------------------------------------
L.C. Karssen
Utrecht
The Netherlands

lennart@karssen.org
http://blog.karssen.org

Stuur mij aub geen Word of Powerpoint bestanden!
Zie http://www.gnu.org/philosophy/no-word-attachments.nl.html
------------------------------------------------------------------

Attachment: signature.asc
Description: OpenPGP digital signature


Reply to: