From mboxrd@z Thu Jan  1 00:00:00 1970
From: john at keeping.me.uk (John Keeping)
Date: Sun, 16 Aug 2015 13:19:54 +0100
Subject: [PATCH 2/3] refactor get_mimetype_from_file() to
 get_mimetype_from_qrypath()
In-Reply-To: <1439669303-17319-1-git-send-email-list@eworm.de>
References: <20150815113043.GD1412@serenity.lan>
 <1439669303-17319-1-git-send-email-list@eworm.de>
Message-ID: <20150816121954.GE1412@serenity.lan>

On Sat, Aug 15, 2015 at 10:08:22PM +0200, Christian Hesse wrote:
> From: Christian Hesse <mail at eworm.de>
> 
> * handle mimetype within a single function
> + return allocated memory on success
> 
> Signed-off-by: Christian Hesse <mail at eworm.de>
> ---
>  cgit.h     |  2 +-
>  shared.c   | 65 +++++++++++++++++++++++++++++++++++---------------------------
>  ui-plain.c | 29 +++++++---------------------
>  3 files changed, 45 insertions(+), 51 deletions(-)
> 
> diff --git a/cgit.h b/cgit.h
> index 0f1e186..d34d353 100644
> --- a/cgit.h
> +++ b/cgit.h
> @@ -391,6 +391,6 @@ extern int readfile(const char *path, char **buf, size_t *size);
>  
>  extern char *expand_macros(const char *txt);
>  
> -extern char *get_mimetype_from_file(const char *filename, const char *ext);
> +extern char *get_mimetype_for_qrypath(const char *qrypath);
>  
>  #endif /* CGIT_H */
> diff --git a/shared.c b/shared.c
> index 5a000a6..0f1f8de 100644
> --- a/shared.c
> +++ b/shared.c
> @@ -561,43 +561,52 @@ char *expand_macros(const char *txt)
>  	return result;
>  }
>  
> -char *get_mimetype_from_file(const char *filename, const char *ext)
> +char *get_mimetype_for_qrypath(const char *qrypath)

If this is called get_mimetype_for_qrypath(), shouldn't it just use
"ctx.qry.path" without taking an argument?  Personally, I'd prefer
get_mimetype_for_filename() taking a "path" argument (also see below
about "path" vs. "ctx.qry.path").

>  {
>  	static const char *delimiters;
> -	char *result;
> -	FILE *fd;
> +	char *ext = NULL, *iterate, *mimetype = NULL, *token;
>  	char line[1024];
> -	char *mimetype;
> -	char *token;
> -
> -	if (!filename)
> -		return NULL;
> +	FILE *fd;
> +	struct string_list_item *mime;
>  
> -	fd = fopen(filename, "r");
> -	if (!fd)
> +	if (qrypath == NULL)
>  		return NULL;
>  
> -	delimiters = " \t\r\n";
> -	result = NULL;
> -
> -	/* loop over all lines in the file */
> -	while (!result && fgets(line, sizeof(line), fd)) {
> -		mimetype = strtok(line, delimiters);
> -
> -		/* skip empty lines and comment lines */
> -		if (!mimetype || (mimetype[0] == '#'))
> -			continue;
> -
> -		/* loop over all extensions of mimetype */
> -		while ((token = strtok(NULL, delimiters))) {
> -			if (!strcasecmp(ext, token)) {
> -				result = xstrdup(mimetype);
> -				break;
> +	ext = strrchr(qrypath, '.');
> +
> +	if (ext && *(++ext)) {
> +		mime = string_list_lookup(&ctx.cfg.mimetypes, ext);
> +		if (mime) {
> +			/* We could just pass the pointer here, but would have to care
> +			 * whether or not to free the memory. Instead just dup. */
> +			mimetype = xstrdup(mime->util);
> +		} else {

Should this be:

		} else if (ctx.cfg.mimetype_file) {

?  Above there's a deleted check on "filename" in the original
get_mimetype_from_file() function.

> +			fd = fopen(ctx.cfg.mimetype_file, "r");
> +			if (fd == NULL)
> +				return NULL;
> +
> +			delimiters = " \t\r\n";
> +
> +			/* loop over all lines in the file */
> +			while (mimetype == NULL && fgets(line, sizeof(line), fd)) {
> +				iterate = strtok(line, delimiters);
> +
> +				/* skip empty lines and comment lines */
> +				if (iterate == NULL || (iterate[0] == '#'))
> +					continue;
> +
> +				/* loop over all extensions of mimetype */
> +				while ((token = strtok(NULL, delimiters))) {
> +					if (strcasecmp(ext, token) == 0) {
> +						mimetype = xstrdup(iterate);
> +						break;
> +					}
> +				}
>  			}
> +			fclose(fd);
>  		}
>  	}
> -	fclose(fd);
>  
> -	return result;
> +	return mimetype;
>  }
>  
> diff --git a/ui-plain.c b/ui-plain.c
> index d68518e..ce59415 100644
> --- a/ui-plain.c
> +++ b/ui-plain.c
> @@ -19,10 +19,8 @@ struct walk_tree_context {
>  static int print_object(const unsigned char *sha1, const char *path)
>  {
>  	enum object_type type;
> -	char *buf, *ext;
> +	char *buf, *mimetype;
>  	unsigned long size;
> -	struct string_list_item *mime;
> -	int freemime;
>  
>  	type = sha1_object_info(sha1, &size);
>  	if (type == OBJ_BAD) {
> @@ -35,22 +33,10 @@ static int print_object(const unsigned char *sha1, const char *path)
>  		cgit_print_error_page(404, "Not found", "Not found");
>  		return 0;
>  	}
> -	ctx.page.mimetype = NULL;
> -	ext = strrchr(path, '.');
> -	freemime = 0;
> -	if (ext && *(++ext)) {
> -		mime = string_list_lookup(&ctx.cfg.mimetypes, ext);
> -		if (mime) {
> -			ctx.page.mimetype = (char *)mime->util;
> -			ctx.page.charset = NULL;
> -		} else {
> -			ctx.page.mimetype = get_mimetype_from_file(ctx.cfg.mimetype_file, ext);
> -			if (ctx.page.mimetype) {
> -				freemime = 1;
> -				ctx.page.charset = NULL;
> -			}
> -		}
> -	}
> +
> +	mimetype = get_mimetype_for_qrypath(ctx.qry.path);

This was "path" before, not "ctx.qry.path".  It looks like they will
always be the same, but we're guaranteed that "path" is non-null.

> +	ctx.page.mimetype = mimetype;
> +
>  	if (!ctx.page.mimetype) {
>  		if (buffer_is_binary(buf, size)) {
>  			ctx.page.mimetype = "application/octet-stream";
> @@ -64,9 +50,8 @@ static int print_object(const unsigned char *sha1, const char *path)
>  	ctx.page.etag = sha1_to_hex(sha1);
>  	cgit_print_http_headers();
>  	html_raw(buf, size);
> -	/* If we allocated this, then casting away const is safe. */
> -	if (freemime)
> -		free((char*) ctx.page.mimetype);
> +	if (mimetype)

There's no need to check "mimetype" before calling free(3), since
free(3) is defined to be a no-op if it's passed a null pointer.

> +		free(mimetype);
>  	return 1;
>  }
>  
> -- 
> 2.5.0