Ecc style errors when running EDK2 CI locally without specifying a target (NO-TARGET)

Ecc style errors when running EDK2 CI locally without specifying a target (NO-TARGET)

[Rebecca Cran]

Is it expected that the EDK2 CI run without errors in NO-TARGET mode - 
that is, when "-t" isn't specified?

When I run it locally with "stuart_ci_build -c .pytool/CISettings.py 
TOOL_CHAIN_TAG=GCC5 -a X64,AARCH64" I get lots of errors: one in 
ArmVirtPkg/PrePi/PrePi.c about a missing comment above a function, and 
then loads in DynamicTablesPkg - e.g.:

PROGRESS - --Running ArmVirtPkg: EccCheck Test NO-TARGET --

ERROR -
ERROR - EFI coding style error
ERROR - *Error code: 8003
ERROR - *The #ifndef at the start of an include file should use both 
prefix and postfix underscore characters, '_'
ERROR - *file: .../edk2/DynamicTablesPkg/Include/Library/AmlLib/AmlLib.h
ERROR - *Line number: 41
ERROR - *The #ifndef name [AML_HANDLE] does not follow the rules
ERROR -


When I specify "-t DEBUG,RELEASE,NOOPT" it completes successfully.


-- 
Rebecca Cran

Re: [edk2-devel] Ecc style errors when running EDK2 CI locally without specifying a target (NO-TARGET)

[Sean]

Not sure i follow your first question but running without a "-t" 
specified at the command line and "-t NO-TARGET" is two different 
things.  For edk2 ci you can see the CISettings.py file defines targets 
supported as a tuple of all 4 ("DEBUG", "RELEASE", "NO-TARGET", "NOOPT").
https://github.com/tianocore/edk2/blob/master/.pytool/CISettings.py#L72

and you can see that if the cli parameter for target is empty then the 
value is assigned to whatever is declared as supported.
https://github.com/tianocore/edk2-pytool-extensions/blob/master/edk2toolext/invocables/edk2_multipkg_aware_invocable.py#L121

By creating a specific target "NO-TARGET" we could easily enable/disable 
their execution and thus we could control it in the matrix on the CI 
servers.  Example here: 
https://github.com/tianocore/edk2/blob/master/.azurepipelines/templates/pr-gate-build-job.yml#L32 



Most of the static code tests run for NO-TARGET.  ECC is such a test so 
it only runs when NO-TARGET is run.

Finally please be aware EccCheck runs git commands and can change your 
local workspace.  This was something i recently learned and should 
probably be revisited as it means running it locally can be destructive 
to changes you don't yet have committed.

Hope that helps.

Thanks
Sean







On 9/30/2020 2:59 PM, Rebecca Cran wrote:
> Is it expected that the EDK2 CI run without errors in NO-TARGET mode - 
> that is, when "-t" isn't specified?
> 
> When I run it locally with "stuart_ci_build -c .pytool/CISettings.py 
> TOOL_CHAIN_TAG=GCC5 -a X64,AARCH64" I get lots of errors: one in 
> ArmVirtPkg/PrePi/PrePi.c about a missing comment above a function, and 
> then loads in DynamicTablesPkg - e.g.:
> 
> PROGRESS - --Running ArmVirtPkg: EccCheck Test NO-TARGET --
> 
> ERROR -
> ERROR - EFI coding style error
> ERROR - *Error code: 8003
> ERROR - *The #ifndef at the start of an include file should use both 
> prefix and postfix underscore characters, '_'
> ERROR - *file: .../edk2/DynamicTablesPkg/Include/Library/AmlLib/AmlLib.h
> ERROR - *Line number: 41
> ERROR - *The #ifndef name [AML_HANDLE] does not follow the rules
> ERROR -
> 
> 
> When I specify "-t DEBUG,RELEASE,NOOPT" it completes successfully.
> 
> 

Re: [EXTERNAL] Re: [edk2-devel] Ecc style errors when running EDK2 CI locally without specifying a target (NO-TARGET)

[Rebecca Cran]

Ah, thanks. I made a wrong assumption about how it works. So I can get 
it to fail faster by running "stuart_ci_build -c .pytool/CISettings.py 
TOOL_CHAIN_TAG=GCC5 -t NO-TARGET" - i.e. running NO-TARGET without also 
running DEBUG,RELEASE,NOOPT too.


It turns out that because of the repo I'm running it against, where I 
haven't kept `master` up-to-date, it thinks almost all the files have 
changed - and so it runs Ecc against code that hasn't really changed for 
a long time.

"INFO - Cmd to run is: git diff --name-status HEAD origin/master"


So, now I've updated my repo's `master`, it both runs much faster, and 
succeeds.

A weekly build which runs Ecc over all the code and compiles a report of 
issues might be nice.


-- 
Rebecca Cran


On 9/30/20 5:59 PM, Sean Brogan wrote:
> Not sure i follow your first question but running without a "-t" 
> specified at the command line and "-t NO-TARGET" is two different 
> things.  For edk2 ci you can see the CISettings.py file defines 
> targets supported as a tuple of all 4 ("DEBUG", "RELEASE", 
> "NO-TARGET", "NOOPT").
> https://github.com/tianocore/edk2/blob/master/.pytool/CISettings.py#L72
>
> and you can see that if the cli parameter for target is empty then the 
> value is assigned to whatever is declared as supported.
> https://github.com/tianocore/edk2-pytool-extensions/blob/master/edk2toolext/invocables/edk2_multipkg_aware_invocable.py#L121 
>
>
> By creating a specific target "NO-TARGET" we could easily 
> enable/disable their execution and thus we could control it in the 
> matrix on the CI servers.  Example here: 
> https://github.com/tianocore/edk2/blob/master/.azurepipelines/templates/pr-gate-build-job.yml#L32 
>
>
>
> Most of the static code tests run for NO-TARGET.  ECC is such a test 
> so it only runs when NO-TARGET is run.
>
> Finally please be aware EccCheck runs git commands and can change your 
> local workspace.  This was something i recently learned and should 
> probably be revisited as it means running it locally can be 
> destructive to changes you don't yet have committed.
>
> Hope that helps.
>
> Thanks
> Sean
>
>
>
>
>
>
>
> On 9/30/2020 2:59 PM, Rebecca Cran wrote:
>> Is it expected that the EDK2 CI run without errors in NO-TARGET mode 
>> - that is, when "-t" isn't specified?
>>
>> When I run it locally with "stuart_ci_build -c .pytool/CISettings.py 
>> TOOL_CHAIN_TAG=GCC5 -a X64,AARCH64" I get lots of errors: one in 
>> ArmVirtPkg/PrePi/PrePi.c about a missing comment above a function, 
>> and then loads in DynamicTablesPkg - e.g.:
>>
>> PROGRESS - --Running ArmVirtPkg: EccCheck Test NO-TARGET --
>>
>> ERROR -
>> ERROR - EFI coding style error
>> ERROR - *Error code: 8003
>> ERROR - *The #ifndef at the start of an include file should use both 
>> prefix and postfix underscore characters, '_'
>> ERROR - *file: .../edk2/DynamicTablesPkg/Include/Library/AmlLib/AmlLib.h
>> ERROR - *Line number: 41
>> ERROR - *The #ifndef name [AML_HANDLE] does not follow the rules
>> ERROR -
>>
>>
>> When I specify "-t DEBUG,RELEASE,NOOPT" it completes successfully.
>>
>>

Re: [EXTERNAL] Re: [edk2-devel] Ecc style errors when running EDK2 CI locally without specifying a target (NO-TARGET)

[Laszlo Ersek]

On 10/01/20 04:25, Rebecca Cran wrote:
> Ah, thanks. I made a wrong assumption about how it works. So I can get
> it to fail faster by running "stuart_ci_build -c .pytool/CISettings.py
> TOOL_CHAIN_TAG=GCC5 -t NO-TARGET" - i.e. running NO-TARGET without also
> running DEBUG,RELEASE,NOOPT too.
> 
> 
> It turns out that because of the repo I'm running it against, where I
> haven't kept `master` up-to-date, it thinks almost all the files have
> changed - and so it runs Ecc against code that hasn't really changed for
> a long time.
> 
> "INFO - Cmd to run is: git diff --name-status HEAD origin/master"
> 
> 
> So, now I've updated my repo's `master`, it both runs much faster, and
> succeeds.

Right; a few days ago I also had to look up the "origin/master"
references in ".pytool/Plugin/EccCheck/EccCheck.py" to understand how it
restricts the warnings to "new" code. I have a dedicated VM for running
CI locally, so whenever I push a topic branch to that, I just have to
push my most recently fetched master branch reference to the VM as well.

Thanks!
Laszlo

> 
> A weekly build which runs Ecc over all the code and compiles a report of
> issues might be nice.
> 
>