git/list[1] front-page[2] threads[3] people[4] search[5] about
 

Re: [PATCH 1/6] do not discard const: the simple cases

From
Junio C Hamano <gitster@pobox.com>
Date
Mar 26, 2026, 17:45 UTC
Message-ID
<xmqqy0jepjqy.fsf@gitster.g>
In-Reply-To
<20260326173402.GB2447148@coredump.intra.peff.net>
Jeff King <peff@peff.net> writes:
Show 35 quoted lines
> On Thu, Mar 26, 2026 at 04:22:47PM +0100, Michael J Gruber wrote:
>
>> This patch covers the easy cases where we deal with a non-const pointer
>> to begin with. It is solved by the cast `bar = (char *) foo`.
>
> I think we can often do better, though. For example, in this case:
>
>> diff --git a/builtin/config.c b/builtin/config.c
>> index 7c4857be62..bd277e5911 100644
>> --- a/builtin/config.c
>> +++ b/builtin/config.c
>> @@ -852,7 +852,7 @@ static int get_urlmatch(const struct config_location_options *opts,
>>  		die("%s", config.url.err);
>>  
>>  	config.section = xstrdup_tolower(var);
>> -	section_tail = strchr(config.section, '.');
>> +	section_tail = (char *) strchr(config.section, '.');
>>  	if (section_tail) {
>>  		*section_tail = '\0';
>>  		config.key = section_tail + 1;
>
> We know that it is OK to cast away the const-ness because config.section
> is writeable, which we know because it just came from xstrdup(). So why
> is it const in the first place? Because the pointer is in a struct which
> may be used with other const strings.
>
> But we can untangle this for the compiler without having to cast by
> using a non-const alias, like:
>
>   char *section;
>   ...
>   config.section = section = xstrdup_tolower(var);
>   section_tail = strchr(section, '.');
>
> Which I think is safer and shows the intent more clearly.
Yeah, this is much clearer.
Previous: Jeff KingNext: Jeff King
Message 17 of 24 in “ISOC23: quell warnings on discarding const”
  1. 0/6 ISOC23: quell warnings on discarding constMichael J Gruber, Mar 26, 2026
  2. 5/6 do not discard const: keep signatureMichael J Gruber, Mar 26, 2026
  3. Junio C HamanoMar 26, 2026
  4. 6/6 do not discard const: the ugly truthMichael J Gruber, Mar 26, 2026
  5. Junio C HamanoMar 26, 2026
  6. Jeff KingMar 26, 2026
  7. 0/4 fix const issues in revision parserJeff King, Mar 26, 2026
  8. 1/4 revision: make handle_dotdot() interface less confusingJeff King, Mar 26, 2026
  9. Junio C HamanoMar 26, 2026
  10. Jeff KingMar 26, 2026
  11. Junio C HamanoMar 27, 2026
  12. 2/4 rev-parse: simplify dotdot parsingJeff King, Mar 26, 2026
  13. 3/4 revision: avoid writing to const string for parent marksJeff King, Mar 26, 2026
  14. 4/4 rev-parse: avoid writing to const string for parent marksJeff King, Mar 26, 2026
  15. 1/6 do not discard const: the simple casesMichael J Gruber, Mar 26, 2026
  16. Jeff KingMar 26, 2026
  17. Junio C HamanoMar 26, 2026
  18. config: store allocated string in non-const pointerJeff King, Mar 26, 2026
  19. 4/6 do not discard const: declare const where we stay constMichael J Gruber, Mar 26, 2026
  20. 2/6 do not discard const: make git-compat-util ISOC23-likeMichael J Gruber, Mar 26, 2026
  21. 3/6 do not discard const: adjust to non-const data typesMichael J Gruber, Mar 26, 2026
  22. Junio C HamanoMar 26, 2026
  23. D. Ben KnobleMar 26, 2026
  24. Michael J GruberMar 27, 2026

Read the whole thread, see it on lore, or plain text.

$ cat FOOTERMessages come from the public archive at lore.kernel.org/git, fetched every hour. The front page is chosen and written each morning by an AI editor and can be wrong; the threads themselves are the record. About and API. For agents: an MCP server at https://gitlist.dev/mcp, and any thread, story or person page as Markdown by adding .md to its URL (or sending Accept: text/markdown). Details in /llms.txt.