[PATCH 3/9] sb: Do not report current date

Joel Sherrill joel.sherrill at oarcorp.com
Tue Dec 9 00:06:50 UTC 2014



On December 8, 2014 6:00:06 PM CST, Peter Dufault <dufault at hda.com> wrote:
>I don't understand what drives this, maybe Sebastian can comment.
>
>I agree with Chris that requiring a documented "grep -v" in an expected
>output script would assist in reproducibility and process validation.
>
>I also agree with Sebastian that "cmp" returning that files are
>identical is very reassuring.

A blind cmp doesn't let the report include host, date, tool versions, etc. 

>> On Dec 8, 2014, at 16:42 , Chris Johns <chrisj at rtems.org> wrote:
>> 
>> On 9/12/2014 8:07 am, Gedare Bloom wrote:
>>> On Mon, Dec 8, 2014 at 3:52 PM, Chris Johns <chrisj at rtems.org>
>wrote:
>>>> On 8/12/2014 5:48 pm, Sebastian Huber wrote:
>>>>> 
>>>>> This makes the report reproducible.
>>>> 
>>>> 
>>>> I think the report should include a date. I do not see any
>advantage having
>>>> reproducible reports. The report captures the specific instance of
>the
>>>> build.
>>>> 
>>> Would it make sense to re-build on a different date and want to
>>> compare the results to see there is no difference?
>> 
>> What you build on a different date cannot be the same by definition. 
>> The date has changed. In a quality context if you reference the first
>
>> build that is what you have. You cannot reference an initial build
>and 
>> then say you used a subsequent build because you know it is the same.
>
>> Where is the dated report to say they are same ? The report is about 
>> reporting what you did and what happened.
>> 
>>> Maybe a flag can be turned on/off for "reproducible" builds.
>> 
>> I do not like flags being available for things like this. The user
>then 
>> needs to audit the setting and this moves the compliance back up to
>the 
>> user.
>> 
>>> Or is it the user's
>>> responsibility to strip out such non-reproducible bits if they want
>>> such a feature?
>> 
>> I can understand an MD5 hash on the components built and that result 
>> being in a report. I have never tried to see if a repeat build of the
>
>> tools produces an exact binary image.
>> 
>> I can also understand a user explicitly adding an exemption to an
>audit 
>> process not to check the report. For example it is common to see
>target 
>> binary images have exemptions for date and time strings embedded in
>them 
>> and a manual audit with a hex dump to verify this is the only
>difference.
>> 
>> Chris
>> _______________________________________________
>> devel mailing list
>> devel at rtems.org
>> http://lists.rtems.org/mailman/listinfo/devel
>
>Peter
>-----------------
>Peter Dufault
>HD Associates, Inc.      Software and System Engineering
>
>_______________________________________________
>devel mailing list
>devel at rtems.org
>http://lists.rtems.org/mailman/listinfo/devel

--joel



More information about the devel mailing list