libs/libc/unistd: Make optarg, optind, opterr and optopt variables.

[casaroli]

Summary

POSIX specifies optarg, optind, opterr and optopt as variables. NuttX defines them as macros over per-task storage, because the tasks of a flat build share one copy of the C library data.

A program that brings its own getopt() and defines these variables does not compile: char *optarg = NULL; expands to char *(*(getoptargp())) = NULL;. A program that only declares them reads a different optarg than its getopt() writes.

In the user space of a kernel build every process has its own copy of the C library data, so there this PR makes them plain variables. getopt_common() copies them to and from its internal state on each call, and getoptargp() and the other accessors return their addresses, so programs built against the old macros keep working. The variables are in a file of their own, so a program that defines them gets no second definition from the library. Flat and protected builds do not change.

The first commit fixes the old nxstyle errors in lib_getopt_common.c (two comments moved above their case labels, one blank line), with no code change: the object code is identical before and after (qemu-armv8a:nsh).

Impact

Kernel builds only: optarg, optind, opterr and optopt are variables in user space. Programs that use them through the macros, or call getopt() from the C library, work as before. Flat and protected builds: no change.

Testing

qemu-armv8a:knsh in QEMU 11. The test programs were built outside the tree against the export package.

A program that defines the four variables and calls getopt(), run as optvars -f file -v x:

Before:

optvars.c:6:1: error: function 'getoptargp' is initialized like a variable

After:

optvars: option f arg file
optvars: option v arg file
optvars: optind 4

The same program without its own definitions gives the same output before and after.

Kernel builds with -Wno-cpp -Werror, each followed by make export and the apps make import (all apps of the configuration compiled for user space): qemu-armv8a:knsh, imx93-evk:knsh, qemu-armv7a:knsh, sabre-6quad:netknsh, sama5d4-ek:knsh, rv-virt:knsh, rv-virt:knsh64, rv-virt:ksmp64, rv-virt:knsh_paging, rv-virt:knetnsh64, rv-virt:kfb64, icicle:knsh, canmv230:knsh, esp32s3-devkit:kernel_n8r2.

Flat builds with -Wno-cpp -Werror, size unchanged: lm3s6965-ek:qemu-flat, mps2-an500:nsh.

tools/checkpatch.sh -c -u -m -g passes.

[xiaoxiang781216]

kernel/flat should keep the same behavior. anyway, we have a similar and better fix, will provide soon.

[casaroli]

OK, please link your PR here when it is ready and I will close this one. I can test it with a program that defines optarg itself.

[github-actions]

MemBrowse Memory Report

esp32-devkitc

• ROM: .flash.text +8 B (+0.0%, 125,096 B / 4,194,272 B, total: 3% used)
• irom0_0_seg: .flash.text +8 B (+0.0%, 89,300 B / 3,342,304 B, total: 3% used)

hifive1-revb

• flash: .text +16 B (+0.0%, 84,252 B / 4,194,304 B, total: 2% used)

qemu-armv8a

• Code: .rodata +8 B, .text.gmtime_r +12 B, .text.init_builtin_run +60 B, .text.work_qqueue +24 B (+0.0%, 347,480 B)

qemu-intel64

• Code: .text +84 B (+0.0%, 8,682,514 B)

rx65n-rsk2mb

• ROM: .rodata +4 B (+0.0%, 87,044 B / 2,097,152 B, total: 4% used)

s698pm-dkit

• Code: .text +592 B (+0.2%, 373,840 B)
  No memory changes detected for:
• arduino-mega2560
• mirtoo
• stm32-nucleo-f103rb