Skip to contentSkip to conversation

docs/misra: document gcc-specific behavior with shifting signed integers

PATCH Superseded Stefano Stabellini 2 participants 3 messages Aug 19, 2023 – Aug 22, 2023

v1 of 2 Superseded Latest is v2 →

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 docs/misra: document gcc-specific behavior with shifting signed integers

From: Stefano Stabellini <[email protected]>

---
Changes in v2:
- use "shift" instead of << or >>
- use All Architectures (I haven't changed all the other instances of x86/arm in the file yet)

---
 docs/misra/C-language-toolchain.rst | 6 ++++++
 1 file changed, 6 insertions(+)
docs/misra/C-language-toolchain.rst +6 -0
index 785aed1eaf..f5ca7bd2c8 100644
@@ -200,6 +200,12 @@ The table columns are as follows:
      - ARM64, X86_64
      - See Section "6.29 Designated Initializers" of GCC_MANUAL
 
+   * - Signed shift acts on negative numbers by sign extension
+     - All architectures
+     - See Section "4.5 Integers" of GCC_MANUAL. As an extension to the
+       C language, GCC does not use the latitude given in C99 and C11
+       only to treat certain aspects of signed shift as undefined.
+
 
 Translation Limits
 __________________
Signature · 1 line
2.25.1
#1 lore raw
Quoted 17 lines · Stefano Stabellini

From: Stefano Stabellini <[email protected]>

---
Changes in v2:
- use "shift" instead of << or >>
- use All Architectures (I haven't changed all the other instances of x86/arm in the file yet)

---
 docs/misra/C-language-toolchain.rst | 6 ++++++
 1 file changed, 6 insertions(+)
docs/misra/C-language-toolchain.rst +0 -0
index 785aed1eaf..f5ca7bd2c8 100644

docs/misra/C-language-toolchain.rst The table columns are as follows: · line 207

Quoted 8 lines
docs/misra/C-language-toolchain.rst +5 -0
@@ -200,6 +200,12 @@ The table columns are as follows:
      - ARM64, X86_64
      - See Section "6.29 Designated Initializers" of GCC_MANUAL
 
+   * - Signed shift acts on negative numbers by sign extension
+     - All architectures
+     - See Section "4.5 Integers" of GCC_MANUAL. As an extension to the
+       C language, GCC does not use the latitude given in C99 and C11
+       only to treat certain aspects of signed shift as undefined.

I'm sorry, but that's still not what the doc says. Replacing << and >> by "shifts" was imo wrong. What's needed instead is that either this is split into two top-level bullet points (one for << and one for >>), or the first sub-bullet-point (which acts as kind of the title) be generalized, with the << and >> details fully moved to the "explanatory" sub-bullet-point.

Jan

#2 lore raw
Quoted 21 lines · Jan Beulich
Quoted 20 lines · Stefano Stabellini

From: Stefano Stabellini <[email protected]>

---
Changes in v2:
- use "shift" instead of << or >>
- use All Architectures (I haven't changed all the other instances of x86/arm in the file yet)

---
 docs/misra/C-language-toolchain.rst | 6 ++++++
 1 file changed, 6 insertions(+)
docs/misra/C-language-toolchain.rst +0 -0
index 785aed1eaf..f5ca7bd2c8 100644
@@ -200,6 +200,12 @@ The table columns are as follows:
      - ARM64, X86_64
      - See Section "6.29 Designated Initializers" of GCC_MANUAL
 
Quoted 11 lines

docs/misra/C-language-toolchain.rst The table columns are as follows: · line 207

Quoted 5 lines
docs/misra/C-language-toolchain.rst +5 -0
+   * - Signed shift acts on negative numbers by sign extension
+     - All architectures
+     - See Section "4.5 Integers" of GCC_MANUAL. As an extension to the
+       C language, GCC does not use the latitude given in C99 and C11
+       only to treat certain aspects of signed shift as undefined.

I'm sorry, but that's still not what the doc says. Replacing << and >> by "shifts" was imo wrong. What's needed instead is that either this is split into two top-level bullet points (one for << and one for >>), or the first sub-bullet-point (which acts as kind of the title) be generalized, with the << and >> details fully moved to the "explanatory" sub-bullet-point.

I think I got your point now