stdc: some improvements
PATCH v2 1/1 stdc: some improvements
- Some small stylistic changes such as removing trailing semicolons and noisy shift calls.
- Ensure pcregrep exists. It is installed as pcre2grep on my machine, so check for both and error out if neither is found. Prefer pcre2grep (installed by libpcre2) because libpcre is EOL.
- Make libc_summ() standard-agnostics by passing in the filter expression and the path to the doc as arguments.
- Make libc_summ() error out if the doc is not found.
- Give basic usage information on usage errors.
- Make the standard version match case insensitive.
---
v1 -> v2:
- Prefer pcre2grep from libpcre2. Suggested by Sam on IRC.
bin/stdc | 101 +++++++++++++++++++++++++++++++++++++------------------ 1 file changed, 68 insertions(+), 33 deletions(-)
bin/stdc +68 -33
@@ -1,65 +1,100 @@ -#!/bin/bash +#!/usr/bin/env bash -set -Eefuo pipefail; +set -Efuo pipefail -prefix="/usr/local"; -datarootdir="$prefix/share"; -docdir="$datarootdir/doc"; +prefix="/usr/local" +datarootdir="$prefix/share" +docdir="$datarootdir/doc" + +c89_filter='/A.3 LIBRARY SUMMARY/,$p' +c89_doc="$docdir/c/c89/c89-draft.txt" +c99_filter='/Library summary$/,/Sequence points$/p' +c99_doc="$docdir/c/c99/n1256.txt" +c11_filter='/Library summary$/,/Sequence points$/p' +c11_doc="$docdir/c/c11/n1570.txt" + +pcregrep="$(type -P pcre2grep)" +if [[ -z "$pcregrep" ]]; then + pcregrep="$(type -P pcregrep)" +fi err() { - >&2 echo "$(basename "$0"): error: $*"; - exit 1; + >&2 echo "$*" } -grep_proto() +die() +{ + err "$(basename "$0"): error: $*" + exit 1 +} + +# Args: +# 1: usage error +usage() { - pcregrep -M "(?s)\b$1 *\([[:alnum:]*,._\s\(\)-]*\);$"; + err "$*" + err + err "usage: $(basename "$0") <version> <function>" + err + err " version ISO C version. Supported versions:" + err " C89, C99, C11" + err " (case insensitive)" + err " function Function to look for" + exit 1 } -libc_summ_c89() +# Args: +# 1: declaration to look for +grep_proto() { - sed -n '/A.3 LIBRARY SUMMARY/,$p' <"$docdir/c/c89/c89-draft.txt"; + "$pcregrep" -M "(?s)\b$1 *\([[:alnum:]*,._\s\(\)-]*\);\$" } +# Args: +# 1: filter expression +# 2: path to plaintext standard libc_summ() { - sed -n '/Library summary$/,/Sequence points$/p'; + if [[ ! -r "$2" ]]; then + die "cannot find or read '$2'" + fi + sed -n -e "$1" "$2" } +if [[ -z "$pcregrep" ]]; then + die "pcre2grep or pcregrep required but is not installed" +fi + case $# in -0) - err "missing ISO C version."; - ;; -1) - err "missing function name."; +0|1) + usage "missing ISO C version and/or function name." ;; 2) ;; *) - shift; - shift; - err "unsupported extra argument(s): $*"; + shift + shift + usage "unsupported extra argument(s): ${*@Q}" ;; -esac; +esac -case "$1" in +case "${1@L}" in c89) - shift; - libc_summ_c89 \ - | grep_proto $@; + libc_summ "$c89_filter" "$c89_doc" \ + | grep_proto "$2" ;; c99) - shift; - libc_summ <"$docdir/c/c99/n1256.txt" \ - | grep_proto $@; + libc_summ "$c99_filter" "$c99_doc" \ + | grep_proto "$2" ;; c11) - shift; - libc_summ <"$docdir/c/c11/n1570.txt" \ - | grep_proto $@; + libc_summ "$c11_filter" "$c11_doc" \ + | grep_proto "$2" ;; *) - err "$1: unsupported ISO C version."; + usage "$1: unsupported ISO C version." ;; -esac; +esac + +# vim: set noexpandtab:
Signature · 1 line
2.39.2
Hi Oskari, Sam,
Quoted 2 lines · Oskari Pirhonen
- Some small stylistic changes such as removing trailing semicolons and noisy
shiftcalls.
But they are sooo pretty =) That's a little OCD of mine putting semicolons in shell scripts. Maybe I read too much C, but it feels good to see punctuation marks at the end of sentences. I prefer to keep semicolons.
Quoted 3 lines
- Ensure pcregrep exists. It is installed as pcre2grep on my machine, so check for both and error out if neither is found. Prefer pcre2grep (installed by libpcre2) because libpcre is EOL.
Hmm, I didn't know about pcre2grep(1). I've applied the following:
commit 40049b4d55f6ba2f392aa3b48835b4555ef79ee2 (HEAD -> main, alx/main)
Author: Alejandro Colomar <[email protected]> Date: Fri Mar 24 12:39:33 2023 +0100
bin/: Use pcre2grep(1) instead of pcregrep(1)
PCRE is EOL, and replaced by PCRE2.
bin/stdc +1 -1
@@ -14,7 +14,7 @@ err() grep_proto() { - pcregrep -M "(?s)\b$1 *\([[:alnum:]*,._\s\(\)-]*\);$"; + pcre2grep -M "(?s)\b$1 *\([[:alnum:]*,._\s\(\)-]*\);$"; } libc_summ_c89()
Quoted 2 lines
- Make
libc_summ()standard-agnostics by passing in the filter expression and the path to the doc as arguments.
It's really standards agnostic except for C89. ISO C drafts are pretty much all the same except for the original one, since I was really ANSI C. That's why C99 and C11 (and C2x if we add it) can reuse libc_summ(), but C89 is an exception.
Quoted 1 line
- Make
libc_summ()error out if the doc is not found.
The shell already errors out with a quite readable message:
$ ./bin/stdc c99 fgets ./bin/stdc: line 54: /usr/local/share/doc/c/c99/n1256.foo: No such file or directory $ echo $? 1
Is it not good enough?
Quoted 1 line
- Give basic usage information on usage errors.
I prefer writing a man page. It keeps the source code simpler. Please keep this suggestion around, and resend it if you feel it's necessary after having a man page (1 or 2 weeks maybe).
Quoted 1 line
- Make the standard version match case insensitive.
Okay.
Quoted 2 lines
Signed-off-by Oskari Pirhonen <[email protected]>
Please send individual patches for each of the logical changes.
Quoted 12 lines
---
v1 -> v2:
- Prefer pcre2grep from libpcre2. Suggested by Sam on IRC.bin/stdc | 101 +++++++++++++++++++++++++++++++++++++------------------ 1 file changed, 68 insertions(+), 33 deletions(-)bin/stdc +0
-0
bin/stdc line 3
Quoted 5 lines
bin/stdc +2
-2@@ -1,65 +1,100 @@ -#!/bin/bash +#!/usr/bin/env bash -set -Eefuo pipefail; +set -Efuo pipefail
Why not set -e? It's not documented in the commit message.
bin/stdc line 12
Quoted 13 lines
bin/stdc +9
-3-prefix="/usr/local"; -datarootdir="$prefix/share"; -docdir="$datarootdir/doc"; +prefix="/usr/local" +datarootdir="$prefix/share" +docdir="$datarootdir/doc" + +c89_filter='/A.3 LIBRARY SUMMARY/,$p' +c89_doc="$docdir/c/c89/c89-draft.txt" +c99_filter='/Library summary$/,/Sequence points$/p' +c99_doc="$docdir/c/c99/n1256.txt" +c11_filter='/Library summary$/,/Sequence points$/p'
This filter is really non_c89_filter :)
bin/stdc line 18
Quoted 6 lines
bin/stdc +6
-0+c11_doc="$docdir/c/c11/n1570.txt" + +pcregrep="$(type -P pcre2grep)" +if [[ -z "$pcregrep" ]]; then + pcregrep="$(type -P pcregrep)" +fi
Not necessary anymore. I switched to pcre2grep after your suggestion. I prefer not adding legacy fallbacks. If someone wants to backport this to an old system, that's not our problem.
Quoted 70 lines
bin/stdc +36
-14err() { - >&2 echo "$(basename "$0"): error: $*"; - exit 1; + >&2 echo "$*" } -grep_proto() +die() +{ + err "$(basename "$0"): error: $*" + exit 1 +} + +# Args: +# 1: usage error +usage() { - pcregrep -M "(?s)\b$1 *\([[:alnum:]*,._\s\(\)-]*\);$"; + err "$*" + err + err "usage: $(basename "$0") <version> <function>" + err + err " version ISO C version. Supported versions:" + err " C89, C99, C11" + err " (case insensitive)" + err " function Function to look for" + exit 1 } -libc_summ_c89() +# Args: +# 1: declaration to look for +grep_proto() { - sed -n '/A.3 LIBRARY SUMMARY/,$p' <"$docdir/c/c89/c89-draft.txt"; + "$pcregrep" -M "(?s)\b$1 *\([[:alnum:]*,._\s\(\)-]*\);\$" } +# Args: +# 1: filter expression +# 2: path to plaintext standard libc_summ() { - sed -n '/Library summary$/,/Sequence points$/p'; + if [[ ! -r "$2" ]]; then + die "cannot find or read '$2'" + fi + sed -n -e "$1" "$2" } +if [[ -z "$pcregrep" ]]; then + die "pcre2grep or pcregrep required but is not installed" +fi + case $# in -0) - err "missing ISO C version."; - ;; -1) - err "missing function name."; +0|1) + usage "missing ISO C version and/or function name." ;; 2) ;; *) - shift; - shift;
bin/stdc line 80
Quoted 10 lines
bin/stdc +5
-3- err "unsupported extra argument(s): $*"; + shift + shift + usage "unsupported extra argument(s): ${*@Q}" ;; -esac; +esac -case "$1" in +case "${1@L}" in
I'm not a fan of shell features. I prefer some command like
"$(printf "%s" "$1" | tr [[:upper:]] [[:lower:]])"
which is much more readable (IMO). Does that look good to you?
bin/stdc line 83
Quoted 6 lines
bin/stdc +2
-3c89) - shift; - libc_summ_c89 \ - | grep_proto $@; + libc_summ "$c89_filter" "$c89_doc" \ + | grep_proto "$2"
I applied the following:
commit 0e54f2590d9689c9b5cef2369d4664f4096fdd37 (HEAD -> main)
Author: Alejandro Colomar <[email protected]> Date: Fri Mar 24 12:56:37 2023 +0100
bin/: Remove some unnecessary shifts and $@s
bin/stdc +3 -6
@@ -45,19 +45,16 @@ esac; case "$1" in c89) - shift; libc_summ_c89 \ - | grep_proto $@; + | grep_proto "$2"; ;; c99) - shift; libc_summ <"$docdir/c/c99/n1256.txt" \ - | grep_proto $@; + | grep_proto "$2"; ;; c11) - shift; libc_summ <"$docdir/c/c11/n1570.txt" \ - | grep_proto $@; + | grep_proto "$2"; ;; *) err "$1: unsupported ISO C version.";
Quoted 13 lines
bin/stdc +3
-6;; c99) - shift; - libc_summ <"$docdir/c/c99/n1256.txt" \ - | grep_proto $@; + libc_summ "$c99_filter" "$c99_doc" \ + | grep_proto "$2" ;; c11) - shift; - libc_summ <"$docdir/c/c11/n1570.txt" \ - | grep_proto $@; + libc_summ "$c11_filter" "$c11_doc" \
bin/stdc line 98
Quoted 10 lines
bin/stdc +5
-2+ | grep_proto "$2" ;; *) - err "$1: unsupported ISO C version."; + usage "$1: unsupported ISO C version." ;; -esac; +esac + +# vim: set noexpandtab:
I'm not sure we want vim comments like this one for every file. Aren't they too noisy? But maybe it's just me. I use things like the following for when I contribute regularly to projects that use a different rule for indentation:
au bufnewfile,bufread ~/src/linux/man-pages/*/man*/* setlocal expandtab au bufnewfile,bufread ~/src/linux/man-pages/*/man*/* setlocal shiftwidth=4 au bufnewfile,bufread ~/src/linux/man-pages/*/man*/* setlocal softtabstop=4 au bufnewfile,bufread ~/src/linux/man-pages/*/man*/* setlocal tabstop=5
au bufnewfile,bufread ~/src/nginx/* setlocal expandtab au bufnewfile,bufread ~/src/nginx/* setlocal shiftwidth=4 au bufnewfile,bufread ~/src/nginx/* setlocal softtabstop=4 au bufnewfile,bufread ~/src/nginx/* setlocal tabstop=5
(tabstop=5 is for realizing when something is actually a tab, which for example happens in Makefiles in NGINX.)
Thanks!
Cheers,
Alex
Signature · 2 lines
<http://www.alejandro-colomar.es/> GPG key fingerprint: A9348594CE31283A826FBDD8D57633D441E25BB5
Quoted 8 lines
Review
Quoted 2 lines
(quoted diff) +2
-0+ +# vim: set noexpandtab:I'm not sure we want vim comments like this one for every file. Aren't they too noisy? But maybe it's just me. I use things like the following for when I contribute regularly to projects that use a different rule for indentation:
I second this. Vim comments pollute source files and are way too editor specific. It's also questionable if they're that useful. If someone sends a patch with spaces instead of tabs, that's easy to catch and fix.
Signature · 2 lines
Best Regards, Tom Schwindl
Hi,
Quoted 11 lines · Alejandro Colomar
Hi Oskari, Sam,
Quoted 2 lines · Oskari Pirhonen
- Some small stylistic changes such as removing trailing semicolons and noisy
shiftcalls.But they are sooo pretty =) That's a little OCD of mine putting semicolons in shell scripts. Maybe I read too much C, but it feels good to see punctuation marks at the end of sentences. I prefer to keep semicolons.
Style is obviously your call here.
Quoted 6 lines
Quoted 3 lines
- Ensure pcregrep exists. It is installed as pcre2grep on my machine, so check for both and error out if neither is found. Prefer pcre2grep (installed by libpcre2) because libpcre is EOL.
Hmm, I didn't know about pcre2grep(1). I've applied the following:
... snip ...
Quoted 11 lines
libc_summ_c89()Quoted 2 lines
- Make
libc_summ()standard-agnostics by passing in the filter expression and the path to the doc as arguments.It's really standards agnostic except for C89. ISO C drafts are pretty much all the same except for the original one, since I was really ANSI C. That's why C99 and C11 (and C2x if we add it) can reuse
libc_summ(), but C89 is an exception.
That's good to know.
Quoted 11 lines
Quoted 1 line
- Make
libc_summ()error out if the doc is not found.The shell already errors out with a quite readable message:
$ ./bin/stdc c99 fgets ./bin/stdc: line 54: /usr/local/share/doc/c/c99/n1256.foo: No such file or directory $ echo $? 1Is it not good enough?
I suppose that is fine for me. I may be too used to having to performing my own checks for files that I ignored Bash being able to do the same.
Quoted 6 lines
Quoted 1 line
- Give basic usage information on usage errors.
I prefer writing a man page. It keeps the source code simpler. Please keep this suggestion around, and resend it if you feel it's necessary after having a man page (1 or 2 weeks maybe).
Heh. Perhaps you're biased towards man pages ;)
I'll keep it around (rebasing as needed for current and possibly future edits).
Quoted 9 lines
Quoted 1 line
- Make the standard version match case insensitive.
Okay.
Quoted 2 lines
Signed-off-by Oskari Pirhonen <[email protected]>Please send individual patches for each of the logical changes.
Whoops. Will do in the future.
Quoted 15 lines
Quoted 15 lines
---
v1 -> v2:
- Prefer pcre2grep from libpcre2. Suggested by Sam on IRC.bin/stdc | 101 +++++++++++++++++++++++++++++++++++++------------------ 1 file changed, 68 insertions(+), 33 deletions(-)bin/stdc +1
-1@@ -1,65 +1,100 @@ -#!/bin/bash +#!/usr/bin/env bash
Quoted 5 lines
bin/stdc line 3
Quoted 2 lines
bin/stdc +1
-1-set -Eefuo pipefail; +set -Efuo pipefailWhy not set -e? It's not documented in the commit message.
I forgot to mention that in the message, but it's no longer relevant since you switched it to just pcre2grep. With set -e it was aborting the script when type -P failed.
... snip ...
Quoted 8 lines
bin/stdc line 4
Quoted 3 lines
bin/stdc +1
-1-case "$1" in +case "${1@L}" inI'm not a fan of shell features. I prefer some command like
"$(printf "%s" "$1" | tr [[:upper:]] [[:lower:]])"which is much more readable (IMO). Does that look good to you?
Fine by me.
Quoted 6 lines
Quoted 6 lines
bin/stdc +2
-3c89) - shift; - libc_summ_c89 \ - | grep_proto $@; + libc_summ "$c89_filter" "$c89_doc" \ + | grep_proto "$2"
... snip ...
Quoted 14 lines
bin/stdc
Quoted 3 lines
bin/stdc +3
-0+esac + +# vim: set noexpandtab:I'm not sure we want vim comments like this one for every file. Aren't they too noisy? But maybe it's just me. I use things like the following for when I contribute regularly to projects that use a different rule for indentation:
au bufnewfile,bufread ~/src/linux/man-pages/*/man*/* setlocal expandtab au bufnewfile,bufread ~/src/linux/man-pages/*/man*/* setlocal shiftwidth=4 au bufnewfile,bufread ~/src/linux/man-pages/*/man*/* setlocal softtabstop=4 au bufnewfile,bufread ~/src/linux/man-pages/*/man*/* setlocal tabstop=5
Quoted 8 lines
au bufnewfile,bufread ~/src/nginx/* setlocal expandtab au bufnewfile,bufread ~/src/nginx/* setlocal shiftwidth=4 au bufnewfile,bufread ~/src/nginx/* setlocal softtabstop=4 au bufnewfile,bufread ~/src/nginx/* setlocal tabstop=5
(tabstop=5 is for realizing when something is actually a tab, which for example happens in Makefiles in NGINX.)
This is an interesting idea that I've never considered before.
Thanks for at least accepting some of the suggestions.
- Oskari