agent.txt

Agent Changes / Reviews

code/code.md:50:5:     Trailing whitespace
code/code.txt:7:8:     delete trailing whitespace
code/code.txt:58:2:    trailing
code/code.txt:100:2:   trailing whitespace
code/code.txt:173:32:  tidy - check files for leading/trailing whitespace at ends of file
code/style2.txt:9:2:   trailing
code/style2.txt:39:34: // comments for single line, or trailing
code/style2.txt:67:2:  trailing punctuation
translate.txt:147:7:   have trailing whitespace -- nuisance for the translator
travis.txt:10:8:       tabs, trailing whitespace

ARRAYs?
attach/attach.h:          FILE **fp_idx;
attach/attach.h:          struct AttachPtr **idx;
attach/attach.h:          struct Body **body_idx;
browser/private_data.h:   char ***files;
complete/data.h:          const char **match_list;
config/mbtable.h:         char **chars;
core/mailbox.h:           struct Email **emails;
core/neomutt.h:           char **env;
email/email.h:            struct Email **emails;
gui/curs_lib.h:           char ***files;
mutt/charset.h:           const char **inrepls;
mutt/hash.h:              struct HashElem **table;
nntp/adata.h:             struct NntpMboxData **groups_list;

certificates
	add relatives to timestamps (X ago, for Y)
	This certificate is valid
	   from Thu, 26 Dec 2024 09:01:01 UTC (x weeks ago)
	     to Wed, 26 Mar 2025 09:01:00 UTC (for y months)
	means lots more translations!
	also key selection

Find more Linked List Nodes, to replace with ARRAY, e.g.
Nodes
	address/group.h:53:2:    STAILQ_HEAD(GroupList,       GroupNode);
	mutt/list.h:41:2:        STAILQ_HEAD(ListHead,        ListNode);
	mutt/observer.h:74:2:    STAILQ_HEAD(ObserverList,    ObserverNode);
	mutt/regex3.h:100:2:     STAILQ_HEAD(RegexList,       RegexNode);
Others
	address/address.h:43:1:   TAILQ_HEAD(AddressList,     Address);
	alias/alias.h:42:1:       TAILQ_HEAD(AliasList,       Alias);
	attach/cid.h:41:2:       STAILQ_HEAD(CidMapList,      CidMap);
	color/attr.h:74:1:        TAILQ_HEAD(AttrColorList,   AttrColor);
	color/curses2.h:48:1:     TAILQ_HEAD(CursesColorList, CursesColor);
	color/regex4.h:47:2:     STAILQ_HEAD(RegexColorList,  RegexColor);
	email/parameter.h:38:1:   TAILQ_HEAD(ParameterList,   Parameter);
	email/tags.h:48:2:       STAILQ_HEAD(TagList,         Tag);
	email/url.h:63:2:        STAILQ_HEAD(UrlQueryList,    UrlQuery);
	gui/mutt_window.h:105:1:  TAILQ_HEAD(MuttWindowList,  MuttWindow);
	hook.c:76:1:              TAILQ_HEAD(HookList,        Hook);
	key/lib.h:76:2:          STAILQ_HEAD(KeymapList,      Keymap);
	mutt/charset.c:80:1:      TAILQ_HEAD(LookupList,      Lookup);
	mutt/logging2.h:88:2:    STAILQ_HEAD(LogLineList,     LogLine);
	mutt/regex3.h:112:2:     STAILQ_HEAD(ReplaceList,     Replace);
	ncrypt/crypt_mod.c:45:2: STAILQ_HEAD(CryptModuleList, CryptModule);

.github README
	welcome

rename *_buffer_*() functions to buf, or strip buffer entirely
char[] -> Buffer conversions
	which?

upstream patches

split up send() to accommodate bg_edit

compose: move Body wrapping into compose

compose: small fns #3222
	<move-to-top>
		simple within a group
		what does it do at the top level?
		altering the first attachment might have consequences
	<move-to-bottom>
		acting on the first attachment might have consequences
	<move-out-of-group>
		moves attached file *after* the group it was in
	<move-into-group>
		moves attached file into the group *above* it

kill the num_accounts and return the value
int mutt_autocrypt_db_account_get_all(struct AutocryptAccount ***accounts, int *num_accounts)

add X_new()/X_free() wrappers for non-trivial objects
	QClass, ListNode
	SmimeKey, PgpUid, MuttThread

free: see neo/[triv/free]

objects must have X_new()/X_free() pair, or neither
	check existing
	51 X_new()
	55 X_free()
	148 struct
	292 [cm]alloc
	find/distinguish between
		struct - simple, data
		object - complex, actions
	data still needs free for ptrs
	too simplistic?

github action
	check for @file header, etc
	*.c need config.h
	*.h need include guard
	check for empty line at eof
	*.c @page
	"calc" undoc'd functions
	shellcheck

eliminate fallthroughs - use functions

why are these 3 charset config not using DT_SLIST?
	they're only allowed *ONE* charset
	mutt_config.c:431:6:  { "charset",            DT_STRING|D_NOT_EMPTY|D_CHARSET_SINGLE, 0,          0, charset_validator,
	mutt_config.c:443:13: { "config_charset",     DT_STRING,                              0,          0, charset_validator,
	nntp/config.c:67:17:  { "newsgroups_charset", DT_STRING,                              IP "utf-8", 0, charset_validator,

bash complete for configure
	push to management
	dev ml

fuzzy matcher

change Score list to use STAILQ

relative time option for the index_format
	3h (ago), 2d, etc

where is punycode encoded/decoded?
	extract to lib and test

need neomutt glossary (search first)
	add to guide

add more auto-completion

security warnings (132)
	30 For loop variable changed in body
	26 Poorly documented large function
	24 Uncontrolled data used in path expression
	21 Local variable address stored in non-local memory
	 9 Year field changed using an arithmetic operation without checking for leap year
	 8 Long switch case
	 3 Nested loops with same variable
	 3 Futile conditional
	 2 Time-of-check time-of-use filesystem race condition
	 2 File created without restricting permissions
	 2 Cleartext transmission of sensitive information
	 1 Uncontrolled process operation
	 1 Unbounded write

turn MoreArgs, MoreArgsF into ONE function and write tests for it
	SKIPWS too?
	generate tests first
	then convert to functions
	then merge functions
find users of all the flags
doc common uses of each flag, e.g. a regex, a file, a colour, a comment, a keybinding, a pattern
suggest better methods for parsing these

merge string/Buffer variants of functions
	(find them all first)

source! (no error if not exist)
	or `source -q`?
		sort of implies that it'll quieten all errors from the source'd file
	or `source -m` missing-allowed
source -d dir (all of dir/*.rc or *.conf)?
	alphabetical / collate
	or `source dir.d` (.d significant) -> find all children or just '*.rc'
muttrc ifdef
	/file absolute
	./file relative source
"source" command relative to $cwd or previous file
source DIR
	=> all children
source command -- allow wildcards
	fits in nicely with ifndef
	or 'source_dir' command to read all contents
	or "source dir" to read all contents
	recursive option?
	recursive by default?

update buf_at() to take negative offsets
	-1 == end of string
	find uses, look for buf_len() or buf->dsize

upgrade buf_seek() to match fseek()
	using SEEK_SET, SEEK_CUR, or SEEK_END (<stdio.h>)
	or our own BUF_ enums

rename *_buffer_*() functions to buf, or strip buffer entirely
rename *_buffer_*() functions to buf, or strip buffer entirely
3	mutt_b64_buffer_decode
5		mutt_b64_decode
4	mutt_b64_buffer_encode
8		mutt_b64_encode
2	mutt_socket_buffer_readln
19		mutt_socket_readln
1	mutt_socket_buffer_readln_d
12		mutt_socket_readln_d
test_mutt_b64_buffer_decode
test_mutt_b64_buffer_encode
find all functions with a Buffer parameter
	some functions might **append** to the dest buffer

refactor "not reached"
	8	ag "not reached"

find potential places for struct Buffer
	691	ag "([a-z_]+(->[a-z_]+)*), sizeof\(\1\)"
	37	ag "([a-z_]+(->[a-z_]+)*)-\>data, \1-\>dsize"

EBNF for all 88 commands?
	categorise the types of all the command parameters
	create a plan for auto-completing them

doxy: next big area for coverage?
	defines 40%
	and drop %age icons?

:changelog ?
	read from /usr/share/doc/neomutt

prep dyk
	need doc repo for saving emails

devel guide: "how to bisect"
	target audience: developer with basic git understanding
	and ability to build neomutt

user guide: "how to bisect config"

add "Add profile picture" to welcome letter
add build and analysis pages to welcome email
	revamp newbie page
collate all my welcome emails
	generate single webpage

repo/org .github README
	web/readme bigger deal about starring/donation
	IPA NeoMutt /ˌniːoʊ/ /mʌt/

Misc
	where are the "source" relative docs?
	alias_file, certificate_file, etc aren't relative
		alias_file
		certificate_file
		debug_file
		history_file
		ssl_ca_certificates_file
	relative to what?
	all *_files relative to:
	1) the source file they're read from
	2) current working directory
	3) User's muttrc? NO
	4) $XDG_CACHE_HOME? NO

how to build neomutt
	w.r.t. neomutt -v
	dependencies

summarise the last year's commits
	then weekly

take my original config docs, plus the diagrams and improve them

∀ command, list:
	where its data is ultimately stored
	who owns the data
	1	Alias Commands (2): alias, unalias
	2	Hook Commands (18): Various hooks like account-hook, folder-hook, send-hook, etc.
	3	Key Binding Commands (6): bind, unbind, macro, unmacro, exec, push
	4	Configuration Commands (10): set, unset, reset, toggle, setenv, unsetenv, etc.
	5	List Management Commands (12): Various list commands and their "un" counterparts
	6	Hook/Pattern Commands (15): Score, spam, color, ignore commands
	7	Mailbox Commands (3): mailboxes, named-mailboxes, unmailboxes
	8	Utility Commands (7): source, echo, cd, version, etc.
	9	Conditional Commands (4): ifdef, ifndef, finish
	10	Other Commands (12): Remaining specialized commands

using `unbind` as a template
	split up all commands
	COMMAND-NAME ->
		parse_command_name()		main handler
		parse_command_name_args()	split line into tokens
			-> struct ParseCommandNameArgs
		parse_command_name_execute()	do the work
	write tests for all _args() functions
		100% coverage
	add _args() to fuzzer?
	might affect parsing
	`mailboxes good1 bad good2; other command`
	current parser fails at 'bad' and stops parsing
	new parser would try to parse until EOL or `;`
		this would allow the option of executing subsequent commands
		probably best to stop on failed command
	arg parser should record where in the statement the arg came from
	error at line X, column Y

tell copilot about ARRAY_FOREACH() when removing items
	and using FREE() which can trample on data

merge
	buf_pretty_mailbox -> mutt_pretty_mailbox
		void buf_pretty_mailbox(struct Buffer *buf)
		void mutt_pretty_mailbox(char *buf, size_t buflen)

	mutt_safe_path -> buf_save_path
		void mutt_safe_path(struct Buffer *dest, const struct Address *a)
		void buf_save_path(struct Buffer *dest, const struct Address *a)
	mutt_save_path
		void mutt_save_path(char *buf, size_t buflen, const struct Address *addr)

apply it to flatcap-brain-dump
	rules: group similar ideas together
	never delete notes
	mark duplicates
	mark similar
	mark obsolete
	leave unknown/ambiguous notes
	apply to `notebook.*` first?
	or tidy existing notes first?
	exclude work-files/ dir
	sort by age
	try to eliminate small files (few lines)
	turn into a mind map!
		Minder (better for beginners)
		Freeplane (better for experts)
	create a word cloud

explain the docs build process to copilot
and get it to suggest something better

apply it to IRC logs
	most asked questions?
	common problems
	common solutions
	suggest help pages

teach copilot about dialogs / functions
	analyse and suggest missing functionality
	differences in behaviour

create a corpus for each fuzzer

create fuzzer for config type parsing

suggest/create dictionary file for fuzzers
	lists of expected keywords

cleanup-by-date .service
	config file or fixed date?

teach copilot about lcov build
	get 100% coverage cases
	lcov --list coverage.info
	Summary
		gcov mutt/file.c
	Details
		gcov -t mutt/file.c (stdout)
		(rather than saving to file.c.gcov)
		-o save-filename
	Functions
		gcov -f mutt/file.c

teach copilot about asan
	env vars
	leak/crash files
	running tests

"error" prefix in error messages
	need refactor + better messages
	post command-parsing refactor

propagate ConfigSubset through libraries
	∀ library, find all Config lookups
	find all library entry points
	make sure ConfigSubset is passed in and propagated

test cases, delete contents of /tmp afterwards
	need list of tests that gen files
	only creates files/dirs in $NEOMUTT_TEST_DIR/tmp/

create a reply flowchart
	who from
	who to
	etc
w.r.t. config, headers, function used
	<group-chat-reply> OP_GROUP_CHAT_REPLY
	<group-reply>      OP_GROUP_REPLY
	<list-reply>       OP_LIST_REPLY
	<reply>            OP_REPLY
list hooks too

teach copilot about release notes
	categorise commits
	reference closed PRs
	collate names
		new / old? how?
how to make a complete release
	today's date
	release notes
	update dates in files
	commit, tag, sign everything
	commit count

hooks: unify logging?
	LL1 `folder-hook PATTERN`
	LL2 command details
	LL1 - broad strokes, user actions
	LL2 - some details

analyse & suggest logging strategy

skill for "refactor parse"
	template code
	naming
	tests
	coverage
	list tokens for fuzzing

skill for tables
	human legibility
	alignment
	clang-format

refactor verbose tests
	pick out input parameters
	pick out output parameters / return code
	turn into a table

vim syntax updating / refatoring