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

Re: [PATCH/RFC v5 7/7] Careful with CRLF when using e.g. UTF-16 for working-tree-encoding

From
Lars Schneider <larsxschneider@gmail.com>
Date
Jan 30, 2018, 11:23 UTC
Message-ID
<55B6C3D5-4131-4636-AD0E-20759EDBE8CD@gmail.com>
In-Reply-To
<20180129201911.9484-1-tboegi@web.de>
Show 7 quoted lines
> On 29 Jan 2018, at 21:19, tboegi@web.de wrote:
> 
> From: Torsten Bögershausen <tboegi@web.de>
> 
> UTF-16 encoded files are treated as "binary" by Git, and no CRLF
> conversion is done.
> When the UTF-16 encoded files are converted into UF-8 using the new
s/UF-8/UTF-8/
Show 12 quoted lines
> "working-tree-encoding", the CRLF are converted if core.autocrlf is true.
> 
> This may lead to confusion:
> A tool writes an UTF-16 encoded file with CRLF.
> The file is commited with core.autocrlf=true, the CLRF are converted into LF.
> The repo is pushed somewhere and cloned by a different user, who has
> decided to use core.autocrlf=false.
> He uses the same tool, and now the CRLF are not there as expected, but LF,
> make the file useless for the tool.
> 
> Avoid this (possible) confusion by ignoring core.autocrlf for all files
> which have "working-tree-encoding" defined.

Maybe I don't understand your use case but I think this will generate even more confusion because that's not what I would expect as a user. I think Git should behave consistently independent of the used encoding. Here are my arguments:

  (1) Legacy users are *not* affected. If you don't use the "working-tree-encoding"
      attribute then nothing changes for you.
  (2) If you use the "working-tree-encoding" attribute *and* you want to ensure 
      your file keeps CRLF then you can define that in the attributes too. E.g.:
      
      *.proj textworking-tree-encoding=UTF-16 eol=crlf
- Lars
Show 71 quoted lines
> The user can still use a .gitattributes file and specify the line endings
> like "text=auto", "text", or "text eol=crlf" and let that .gitattribute
> file travel together with push and clone.
> 
> Change convert.c to e more careful, simplify the initialization when
> attributes are retrived (and none are specified) and update the documentation.
> 
> Signed-off-by: Torsten Bögershausen <tboegi@web.de>
> ---
> Documentation/gitattributes.txt |  9 ++++++---
> convert.c                       | 15 ++++++++++++---
> 2 files changed, 18 insertions(+), 6 deletions(-)
> 
> diff --git a/Documentation/gitattributes.txt b/Documentation/gitattributes.txt
> index a8dbf4be3..3665c4677 100644
> --- a/Documentation/gitattributes.txt
> +++ b/Documentation/gitattributes.txt
> @@ -308,12 +308,15 @@ Use the `working-tree-encoding` attribute only if you cannot store a file in
> UTF-8 encoding and if you want Git to be able to process the content as
> text.
> 
> +Note that when `working-tree-encoding` is defined, core.autocrlf is ignored.
> +Set the `text` attribute (or `text=auto`) to enable CRLF conversions.
> +
> Use the following attributes if your '*.txt' files are UTF-16 encoded
> -with byte order mark (BOM) and you want Git to perform automatic line
> -ending conversion based on your platform.
> +with byte order mark (BOM) and you want Git to perform line
> +ending conversion based on core.eol.
> 
> ------------------------
> -*.txt		text working-tree-encoding=UTF-16
> +*.txt		working-tree-encoding=UTF-16 text
> ------------------------
> 
> Use the following attributes if your '*.txt' files are UTF-16 little
> diff --git a/convert.c b/convert.c
> index 13fad490c..e7f11d1db 100644
> --- a/convert.c
> +++ b/convert.c
> @@ -1264,15 +1264,24 @@ static void convert_attrs(struct conv_attrs *ca, const char *path)
> 		}
> 		ca->checkout_encoding = git_path_check_encoding(ccheck + 5);
> 	} else {
> -		ca->drv = NULL;
> -		ca->crlf_action = CRLF_UNDEFINED;
> -		ca->ident = 0;
> +		memset(ca, 0, sizeof(*ca));
> 	}
> 
> 	/* Save attr and make a decision for action */
> 	ca->attr_action = ca->crlf_action;
> 	if (ca->crlf_action == CRLF_TEXT)
> 		ca->crlf_action = text_eol_is_crlf() ? CRLF_TEXT_CRLF : CRLF_TEXT_INPUT;
> +	/*
> +	 * Often UTF-16 encoded files are read and written by programs which
> +	 * really need CRLF, and it is important to keep the CRLF "as is" when
> +	 * files are committed with core.autocrlf=true and the repo is pushed.
> +	 * The CRLF would be converted into LF when the repo is cloned to
> +	 * a machine with core.autocrlf=false.
> +	 * Obey the "text" and "eol" attributes and be independent on the
> +	 * local core.autocrlf for all "encoded" files.
> +	 */
> +	if ((ca->crlf_action == CRLF_UNDEFINED) && ca->checkout_encoding)
> +		ca->crlf_action = CRLF_BINARY;
> 	if (ca->crlf_action == CRLF_UNDEFINED && auto_crlf == AUTO_CRLF_FALSE)
> 		ca->crlf_action = CRLF_BINARY;
> 	if (ca->crlf_action == CRLF_UNDEFINED && auto_crlf == AUTO_CRLF_TRUE)
> -- 
> 2.16.0.rc0.2.g64d3e4d0cc.dirty
>
Previous: tboegi@web.deNext: Torsten Bögershausen
Message 35 of 43 in “convert: add support for different encodings”
  1. 0/6 convert: add support for different encodingslars.schneider@autodesk.com, Jan 20, 2018
  2. 1/6 strbuf: remove unnecessary NUL assignment in xstrdup_tolower()lars.schneider@autodesk.com, Jan 20, 2018
  3. 2/6 strbuf: add xstrdup_toupper()lars.schneider@autodesk.com, Jan 20, 2018
  4. 3/6 utf8: add function to detect prohibited UTF-16/32 BOMlars.schneider@autodesk.com, Jan 20, 2018
  5. 4/6 utf8: add function to detect a missing UTF-16/32 BOMlars.schneider@autodesk.com, Jan 20, 2018
  6. 5/6 convert: add 'working-tree-encoding' attributelars.schneider@autodesk.com, Jan 20, 2018
  7. Simon RuderichJan 21, 2018
  8. Lars SchneiderJan 22, 2018
  9. Jeff KingJan 23, 2018
  10. Simon RuderichJan 23, 2018
  11. Jeff KingJan 23, 2018
  12. Junio C HamanoJan 23, 2018
  13. Simon RuderichJan 23, 2018
  14. SQUASH convert: add tracing for 'working-tree-encoding' attributelars.schneider@autodesk.com, Jan 22, 2018
  15. Eric SunshineJan 22, 2018
  16. SQUASH convert: add tracing for 'working-tree-encoding' attributelars.schneider@autodesk.com, Jan 23, 2018
  17. 6/6 convert: add tracing for 'working-tree-encoding' attributelars.schneider@autodesk.com, Jan 20, 2018
  18. Torsten BögershausenJan 23, 2018
  19. Junio C HamanoJan 23, 2018
  20. 0/7 convert: add support for different encodingstboegi@web.de, Jan 29, 2018
  21. 2/7 strbuf: add xstrdup_toupper()tboegi@web.de, Jan 29, 2018
  22. 3/7 utf8: add function to detect prohibited UTF-16/32 BOMtboegi@web.de, Jan 29, 2018
  23. 6/7 convert: add tracing for 'working-tree-encoding' attributetboegi@web.de, Jan 29, 2018
  24. 4/7 utf8: add function to detect a missing UTF-16/32 BOMtboegi@web.de, Jan 29, 2018
  25. Junio C HamanoJan 30, 2018
  26. Lars SchneiderJan 30, 2018
  27. Junio C HamanoJan 30, 2018
  28. 5/7 convert: add 'working-tree-encoding' attributetboegi@web.de, Jan 29, 2018
  29. Junio C HamanoJan 30, 2018
  30. Lars SchneiderJan 30, 2018
  31. Junio C HamanoJan 30, 2018
  32. Lars SchneiderJan 31, 2018
  33. Junio C HamanoJan 31, 2018
  34. 7/7 Careful with CRLF when using e.g. UTF-16 for working-tree-encodingtboegi@web.de, Jan 29, 2018
  35. Lars SchneiderJan 30, 2018
  36. Torsten BögershausenJan 30, 2018
  37. Lars SchneiderJan 30, 2018
  38. Torsten BögershausenJan 31, 2018
  39. Lars SchneiderJan 31, 2018
  40. Junio C HamanoFeb 2, 2018
  41. Torsten BögershausenFeb 7, 2018
  42. Junio C HamanoFeb 7, 2018
  43. 1/7 strbuf: remove unnecessary NUL assignment in xstrdup_tolower()tboegi@web.de, Jan 29, 2018

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.