Remove use of STATIC macros from EDK II C Coding Standard Specification (Bugzilla Bug 1766)

[tianocore-issues]

This issue was created automatically with bugzilla2github

Bugzilla Bug 1766

Date: 2019-04-29T12:27:36+00:00
From: @mdkinney
To: @changab
CC: @lgao4, @bexcran, yonghong.zhu

Blocker for: #1767
Last updated: 2023-05-06T05:53:46+00:00

[tianocore-issues]

Comment 7341

Date: 2019-04-29 12:27:36 +0000
From: @mdkinney

• Industry Specification: ---
• Release Observed: EDK II Master
• Releases to Fix: EDK II Master
• Target OS: ---
• Bugzilla Assignee(s): @changab

Update EDK II C Coding Standard Specification to remove the use of STATIC macros. Instead, the standard C syntax of 'static' should be used.

This macro was introduced a long time ago so it could be overloaded to other value to help with debugging. These debugging issues are not an issue anymore with the supported toolchains, so this document should recommend the use of the standard C syntax.

[tianocore-issues]

Comment 7343

Date: 2019-04-29 12:43:47 +0000
From: @mdkinney

Also add recommendations on when to use 'static'.

There is feedback in the email thread above that compilers/linkers can
produce more efficient code if 'static' is used more aggressively.

Another advantage of using 'static' for global variables that are scoped
to a single file is to prevent name collisions between modules and libraries.

Add recommendations to the EDK II C Coding Standards Specification to use
'static' for all functions and global variables that are not referenced
outside the current C file.

[tianocore-issues]

Comment 7453

Date: 2019-05-08 23:04:26 +0000
From: Yonghong Zhu <<yonghong.zhu>>

Mike, please help on it. thanks.

[tianocore-issues]

Comment 14540

Date: 2020-12-09 15:36:56 +0000
From: @bexcran

Has the problem of debuggers not coping with static functions been fixed?

[tianocore-issues]

Comment 14541

Date: 2020-12-09 15:38:40 +0000
From: @bexcran

Using static instead of STATIC would conflict with the rule about not using C types.

[tianocore-issues]

Comment 20038

Date: 2022-11-15 22:08:32 +0000
From: @lgao4

Abner will update it.

[tianocore-issues]

Comment 20062

Date: 2022-11-18 22:26:22 +0000
From: @changab

Remove the entire 5.4.2.2 section. We are not allowed to use upper-case STAIC in the source file now. Just follow C standard and use the lower-case 'static'.

[tianocore-issues]

Comment 21501

Date: 2023-05-06 05:53:46 +0000
From: @bexcran

There are still several locations that need updated. In addition, I seem to recall a decision to allow use of static functions.

5.4.2.2 Static
An object declared STATIC has either file or block scope

5.4.2.2.2 Functions should not be declared STATIC.
Some source-level debuggers are unable to resolve static functions. Until it can be verified that no one
is dependent upon a debugger with this limitation, it is strongly recommended that functions not be
declared static.

5.6.1.2 The use of int, unsigned, char, void, static, long is a violation
of the coding convention.

[plgibbons-lenovo]

I saw this change and, while I am very pleased to see it, why only 'static'? Why not dispense with the other unnecessary re-definitions of standard C keywords: void, const. My understanding for the original rationale was because of EBC, but that seems largely irrelevant now (and I never understood why EBC couldn't parse 'void' as well as 'VOID').
Diverging from standard C just makes it harder to reuse code in multiple contexts, forces new devs to learn yet another system, etc. I think it would be an improvement if EDK2 could become more standard and allow use of more standard modern type names like uint64_t. I realize there have historically been issues with using preexisting compiler libs and headers, but why not adopt standard things when we can? Standardization is good, but not standard non-standardization, which seems like the path Tiano/EDK/EDK2 went down.

[mdkinney]

Resolved in commit 422e975

[mdkinney]

I saw this change and, while I am very pleased to see it, why only 'static'? Why not dispense with the other unnecessary re-definitions of standard C keywords: void, const. My understanding for the original rationale was because of EBC, but that seems largely irrelevant now (and I never understood why EBC couldn't parse 'void' as well as 'VOID'). Diverging from standard C just makes it harder to reuse code in multiple contexts, forces new devs to learn yet another system, etc. I think it would be an improvement if EDK2 could become more standard and allow use of more standard modern type names like uint64_t. I realize there have historically been issues with using preexisting compiler libs and headers, but why not adopt standard things when we can? Standardization is good, but not standard non-standardization, which seems like the path Tiano/EDK/EDK2 went down.

The EDK2 project in TianoCore provides a conformant implementation of the UEFI Specification. The data types are defined in https://uefi.org/specs/UEFI/2.11/02_Overview.html#data-types. This includes CONST, VOID, and the signed and unsigned integer types INTxx, UINTxx, so those can not be removed.

The UEFI Specification does not define STATIC. That was an EDK2 implementation choice to define that keyword. The use case for that define is no longer valid and can be safely removed from EDK II C Coding Standard Specification.

Would require UEFI Specification changes to remove the use of these types.