Re: [edk2-devel] [edk2-platforms][PATCH 1/1] Platforms/RaspberryPi: Fix BIOS Release Date and System Manufacturer

["Andrew Fish" <afish@apple.com>]

[-- Attachment #1: Type: text/plain, Size: 5653 bytes --]



> On Jul 23, 2020, at 8:47 AM, Leif Lindholm <leif@nuviainc.com> wrote:
> 
> Hi Pete,
> 
> On Thu, Jul 23, 2020 at 15:38:55 +0100, Pete Batard wrote:
>>>> @@ -626,6 +627,28 @@ BIOSInfoUpdateSmbiosType0 (
>>>>    INTN   i;
>>>>    INTN   State = 0;
>>>>    INTN   Value[2];
>>>> +  INTN   Year = (__DATE__[7] == '?' ? 1900  \
>>>> +           : (((__DATE__[7] - '0') * 1000 ) \
>>>> +           + (__DATE__[8] - '0') * 100      \
>>>> +           + (__DATE__[9] - '0') * 10       \
>>>> +           + __DATE__[10] - '0'));
>>>> +  INTN   Month = ( __DATE__ [2] == '?' ? 1  \
>>>> +           : __DATE__ [2] == 'n' ? (        \
>>>> +             __DATE__ [1] == 'a' ? 1 : 6)   \
>>>> +           : __DATE__ [2] == 'b' ? 2        \
>>>> +           : __DATE__ [2] == 'r' ? (        \
>>>> +             __DATE__ [0] == 'M' ? 3 : 4)   \
>>>> +           : __DATE__ [2] == 'y' ? 5        \
>>>> +           : __DATE__ [2] == 'l' ? 7        \
>>>> +           : __DATE__ [2] == 'g' ? 8        \
>>>> +           : __DATE__ [2] == 'p' ? 9        \
>>>> +           : __DATE__ [2] == 't' ? 10       \
>>>> +           : __DATE__ [2] == 'v' ? 11       \
>>>> +           : 12);
>>>> +  INTN   Day = ( __DATE__[4] == '?' ? 1     \
>>>> +           : ((__DATE__[4] == ' ' ? 0 :     \
>>>> +             ((__DATE__[4] - '0') * 10))    \
>>>> +           + __DATE__[5] - '0'));
>>> 
>>> So, this hunk is very neat, but nigh-on unreviewable.
>>> I.e. we should defintely have it - but only once.
>>> 
>>> Could you break this up into some macros to go into some generic
>>> helper lib? (I don't have a better idea than EmbeddedPkg TimeBaseLib,
>>> but then that is already included in this module.)
>> 
>> So you would like to have a set of macros like:
>> TIME_GET_BUILD_YEAR, TIME_GET_BUILD_MONTH, TIME_GET_BUILD_DAY in
>> TimeBaseLib.h that perform the above?
>> 
>>> Would you be OK to break that snippet out separate?
>> 
>> I think that's a good idea, especially as there is a potential underlying
>> issue with the __DATE__ format being specific to each compiler, so we
>> probably want to handle compiler detection somewhere, preferably globally.
>> For instance, the Intel compiler's __DATE__ format would not work with the
>> above, so I'll add some "vetted compiler" detection for the macros.
> 
> Yes, thanks.
> A friendly "your compiler probably won't know what to do with this"
> #error would be friendlier than whatever non-GCC/clang would generate
> on encountering this.
> 

Leif,

I was thinking the "Mmm dd yyyy” form of __DATE__ was standard. Maybe all we need is a compile time assert that checks the format?

6.10.8.1 Mandatory macros 1 The following macro names shall be defined by the implementation: _ _DATE_ _ The date of translation of the preprocessing translation unit: a character string literal of the form "Mmm dd yyyy", where the names of the months are the same as those generated by the asctime function, and the first character of dd is a space character if the value is less than 10. If the date of translation is not available, an implementation-defined valid date shall be supplied 

It seems likely some compilers had an alternate version and chose not to change it and break existing code when the format got standardized. You would hope those compilers had some kind of optional strict ANSI flag or some such….

Thanks,

Andrew Fish

>> The one thing I am not planning to look into is that, of course, as long as
>> you don't explicitly force the compiler to rebuild the sources where these
>> macros are used, then you may end up with a very obsolete build date
>> compared to the actual date of your last re-build.
> 
> Yeah, that's fine.
> 
>> But that's mostly because
>> I don't think there exists a generic solution we can ise to force
>> recompilation of a file that uses a specific macro and also because our main
>> goal with these is to ensure that the Pi firmwares, that we produce through
>> AppVeyor, have a proper build date, and AppVeyor builds are are always
>> clean.
> 
> Yeah, I really don't care about that.
> Or the flipside, where someone modifies only (say) the Smbios
> populating code a month after they built the rest of the tree.
> 
> If someone wants to put a *specific* date in, that should be done
> manually, and if someone wants information on what was actually built,
> they should put in the hashes of the top commits of all repositories
> on PACKAGES_PATH as well as whether the state of each was clean.
> 
> Managing these bits through a rigid build process, like you are, is
> very much valid.
> 
>> I'll send an EDK2 patch for the macros, and then a revised patch for this
>> when I get a chance.
> 
> Thanks!
> 
> Regards,
> 
> Leif
> 
>> Regards,
>> 
>> /Pete
>> 
>>> 
>>> /
>>>     Leif
>>> 
>>>>    // Populate the Firmware major and minor.
>>>>    Status = mFwProtocol->GetFirmwareRevision (&EpochSeconds);
>>>> @@ -648,6 +671,10 @@ BIOSInfoUpdateSmbiosType0 (
>>>>      mBiosVendor, sizeof (mBiosVendor));
>>>>    UnicodeStrToAsciiStrS ((CHAR16*)PcdGetPtr (PcdFirmwareVersionString),
>>>>      mBiosVersion, sizeof (mBiosVersion));
>>>> +  ASSERT (Year >= 0 && Year <= 9999);
>>>> +  ASSERT (Month >= 1 && Month <= 12);
>>>> +  ASSERT (Day >= 1 && Day <= 31);
>>>> +  AsciiSPrint (mBiosDate, sizeof (mBiosDate), "%02d/%02d/%04d", Month, Day, Year);
>>>>    // Look for a "x.y" numeric string anywhere in mBiosVersion and
>>>>    // try to parse it to populate the BIOS major and minor.
>>>> -- 
>>>> 2.21.0.windows.1
>>>> 
>> 
> 
> 
> 


[-- Attachment #2: Type: text/html, Size: 10474 bytes --]