Skip to content

Commit ac4d870

Browse files
dschodscho
committed
range-diff: support reading mbox files
Internally, the `git range-diff` command spawns a `git log` process and parses its output for the given commit ranges. This works well when the patches that need to be compared are present in the local repository in the form of commits. In scenarios where that is not the case, the `range-diff` command is currently less helpful. The Git mailing list is such a scenario: Instead of using Git to exchange commits, the patches are sent there as plain-text and no commit range can be specified to let `range-diff` consume those patches. Instead, the expectation is to download the mails, apply them locally and then use `range-diff`. This can be quite cumbersome e.g. when a suitable base revision has to be found first where the patch applies cleanly. Let's offer a way to read those patches from pre-prepared MBox files instead when an argument "mbox:<filename>" is passed instead of a commit range. For extra convenience, interpret the filename `-` as standard input. This makes it easy to compare contributions on the mailing list with the actual commits that were integrated into Git's main branch. Example: commit=5c4003ca3f0e9ac6d3c8aa3e387ff843bd440411 mid=bdfa3845b81531863941e6a97c28eb1afa62dd2c.1489435755.git.johannes.schindelin@gmx.de curl -s https://lore.kernel.org/git/$mid/raw | git range-diff mbox:- $commit^! This addresses #207 Signed-off-by: Johannes Schindelin <[email protected]>
1 parent b757478 commit ac4d870

File tree

3 files changed

+366
-2
lines changed

3 files changed

+366
-2
lines changed

‎Documentation/git-range-diff.txt

Lines changed: 2 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -37,7 +37,8 @@ There are three ways to specify the commit ranges:
3737

3838
- `<range1> <range2>`: Either commit range can be of the form
3939
`<base>..<rev>`, `<rev>^!` or `<rev>^-<n>`. See `SPECIFYING RANGES`
40-
in linkgit:gitrevisions[7] for more details.
40+
in linkgit:gitrevisions[7] for more details. Alternatively, the
41+
patches can be provided as an mbox-formatted file via `mbox:<path>`.
4142

4243
- `<rev1>...<rev2>`. This is equivalent to
4344
`<rev2>..<rev1> <rev1>..<rev2>`.

‎range-diff.c

Lines changed: 337 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -12,6 +12,7 @@
1212
#include "userdiff.h"
1313
#include "apply.h"
1414
#include "revision.h"
15+
#include "dir.h"
1516

1617
struct patch_util {
1718
/* For the search for an exact match */
@@ -26,6 +27,313 @@ struct patch_util {
2627
struct object_id oid;
2728
};
2829

30+
static inline int strtost(char const *s, size_t *result, const char **end)
31+
{
32+
unsigned long u;
33+
char *p;
34+
35+
errno = 0;
36+
/* negative values would be accepted by strtoul */
37+
if (!isdigit(*s))
38+
return -1;
39+
u = strtoul(s, &p, 10);
40+
if (errno || p == s)
41+
return -1;
42+
if (result)
43+
*result = u;
44+
*end = p;
45+
46+
return 0;
47+
}
48+
49+
static int parse_hunk_header(const char *p,
50+
size_t *old_count, size_t *new_count,
51+
const char **end)
52+
{
53+
size_t o = 1, n = 1;
54+
55+
if (!skip_prefix(p, "@@ -", &p) ||
56+
strtost(p, NULL, &p) ||
57+
/* The range is -<start>[,<count>], defaulting to count = 1 */
58+
!(*p == ' ' || (*p == ',' && !strtost(p + 1, &o, &p))) ||
59+
!skip_prefix(p, " +", &p) ||
60+
strtost(p, NULL, &p) ||
61+
/* The range is +<start>[,<count>], defaulting to count = 1 */
62+
!(*p == ' ' || (*p == ',' && !strtost(p + 1, &n, &p))) ||
63+
!skip_prefix(p, " @@", &p))
64+
return -1;
65+
66+
*old_count = o;
67+
*new_count = n;
68+
*end = p;
69+
70+
return 0;
71+
}
72+
73+
/*
74+
* This function finds the end of the line, replaces the newline character with
75+
* a NUL, and returns the offset of the start of the next line.
76+
*
77+
* If no newline character was found, it returns the offset of the trailing NUL
78+
* instead.
79+
*/
80+
static inline int find_next_line(const char *line, size_t size)
81+
{
82+
char *eol;
83+
84+
eol = memchr(line, '\n', size);
85+
if (!eol)
86+
return size;
87+
88+
*eol = '\0';
89+
90+
return eol + 1 - line;
91+
}
92+
93+
static int read_mbox(const char *path, struct string_list *list)
94+
{
95+
struct strbuf buf = STRBUF_INIT, contents = STRBUF_INIT;
96+
struct strbuf long_subject = STRBUF_INIT;
97+
struct patch_util *util = NULL;
98+
enum {
99+
MBOX_BEFORE_HEADER,
100+
MBOX_IN_HEADER,
101+
MBOX_IN_COMMIT_MESSAGE,
102+
MBOX_AFTER_TRIPLE_DASH,
103+
MBOX_IN_DIFF
104+
} state = MBOX_BEFORE_HEADER;
105+
char *line, *current_filename = NULL;
106+
int len;
107+
size_t size, old_count = 0, new_count = 0;
108+
const char *author = NULL, *subject = NULL;
109+
110+
if (!strcmp(path, "-")) {
111+
if (strbuf_read(&contents, STDIN_FILENO, 0) < 0)
112+
return error_errno(_("could not read stdin"));
113+
} else if (strbuf_read_file(&contents, path, 0) < 0)
114+
return error_errno(_("could not read '%s'"), path);
115+
116+
line = contents.buf;
117+
size = contents.len;
118+
for (; size; size -= len, line += len) {
119+
const char *p;
120+
121+
len = find_next_line(line, size);
122+
123+
if (state == MBOX_BEFORE_HEADER) {
124+
parse_from_delimiter:
125+
if (!skip_prefix(line, "From ", &p))
126+
continue;
127+
128+
if (util)
129+
BUG("util already allocated");
130+
util = xcalloc(1, sizeof(*util));
131+
if (get_oid_hex(p, &util->oid) < 0)
132+
oidcpy(&util->oid, null_oid());
133+
util->matching = -1;
134+
author = subject = NULL;
135+
136+
state = MBOX_IN_HEADER;
137+
continue;
138+
}
139+
140+
if (starts_with(line, "diff --git ")) {
141+
struct patch patch = { 0 };
142+
struct strbuf root = STRBUF_INIT;
143+
int linenr = 0;
144+
int orig_len;
145+
146+
state = MBOX_IN_DIFF;
147+
old_count = new_count = 0;
148+
strbuf_addch(&buf, '\n');
149+
if (!util->diff_offset)
150+
util->diff_offset = buf.len;
151+
152+
orig_len = len;
153+
/* `find_next_line()`'s replaced the LF with a NUL */
154+
line[len - 1] = '\n';
155+
len = len > 1 && line[len - 2] == '\r' ?
156+
error(_("cannot handle diff headers with "
157+
"CR/LF line endings")) :
158+
parse_git_diff_header(&root, &linenr, 1, line,
159+
len, size, &patch);
160+
if (len < 0) {
161+
error(_("could not parse git header '%.*s'"),
162+
orig_len, line);
163+
release_patch(&patch);
164+
free(util);
165+
free(current_filename);
166+
string_list_clear(list, 1);
167+
strbuf_release(&buf);
168+
strbuf_release(&contents);
169+
strbuf_release(&long_subject);
170+
return -1;
171+
}
172+
173+
if (patch.old_name)
174+
skip_prefix(patch.old_name, "a/",
175+
(const char **)&patch.old_name);
176+
if (patch.new_name)
177+
skip_prefix(patch.new_name, "b/",
178+
(const char **)&patch.new_name);
179+
180+
strbuf_addstr(&buf, " ## ");
181+
if (patch.is_new)
182+
strbuf_addf(&buf, "%s (new)", patch.new_name);
183+
else if (patch.is_delete)
184+
strbuf_addf(&buf, "%s (deleted)", patch.old_name);
185+
else if (patch.is_rename)
186+
strbuf_addf(&buf, "%s => %s", patch.old_name, patch.new_name);
187+
else
188+
strbuf_addstr(&buf, patch.new_name);
189+
190+
free(current_filename);
191+
if (patch.is_delete)
192+
current_filename = xstrdup(patch.old_name);
193+
else
194+
current_filename = xstrdup(patch.new_name);
195+
196+
if (patch.new_mode && patch.old_mode &&
197+
patch.old_mode != patch.new_mode)
198+
strbuf_addf(&buf, " (mode change %06o => %06o)",
199+
patch.old_mode, patch.new_mode);
200+
201+
strbuf_addstr(&buf, " ##\n");
202+
util->diffsize++;
203+
release_patch(&patch);
204+
} else if (state == MBOX_IN_HEADER) {
205+
if (!line[0]) {
206+
state = MBOX_IN_COMMIT_MESSAGE;
207+
/* Look for an in-body From: */
208+
if (skip_prefix(line + 1, "From: ", &p)) {
209+
size -= p - line;
210+
line += p - line;
211+
len = find_next_line(line, size);
212+
213+
while (isspace(*p))
214+
p++;
215+
author = p;
216+
}
217+
strbuf_addstr(&buf, " ## Metadata ##\n");
218+
if (author)
219+
strbuf_addf(&buf, "Author: %s\n", author);
220+
strbuf_addstr(&buf, "\n ## Commit message ##\n");
221+
if (subject)
222+
strbuf_addf(&buf, " %s\n\n", subject);
223+
} else if (skip_prefix(line, "From: ", &p)) {
224+
while (isspace(*p))
225+
p++;
226+
author = p;
227+
} else if (skip_prefix(line, "Subject: ", &p)) {
228+
const char *q;
229+
230+
while (isspace(*p))
231+
p++;
232+
subject = p;
233+
234+
if (starts_with(p, "[PATCH") &&
235+
(q = strchr(p, ']'))) {
236+
q++;
237+
while (isspace(*q))
238+
q++;
239+
subject = q;
240+
}
241+
242+
if (len < size && line[len] == ' ') {
243+
/* handle long subject */
244+
strbuf_reset(&long_subject);
245+
strbuf_addstr(&long_subject, subject);
246+
while (len < size && line[len] == ' ') {
247+
line += len;
248+
size -= len;
249+
len = find_next_line(line, size);
250+
strbuf_addstr(&long_subject, line);
251+
}
252+
subject = long_subject.buf;
253+
}
254+
}
255+
} else if (state == MBOX_IN_COMMIT_MESSAGE) {
256+
if (!line[0]) {
257+
strbuf_addch(&buf, '\n');
258+
} else if (strcmp(line, "---")) {
259+
int tabs = 0;
260+
261+
/* simulate tab expansion */
262+
while (line[tabs] == '\t')
263+
tabs++;
264+
strbuf_addf(&buf, "%*s%s\n",
265+
4 + 8 * tabs, "", line + tabs);
266+
} else {
267+
/*
268+
* Trim the trailing newline that is added
269+
* by `format-patch`.
270+
*/
271+
strbuf_trim_trailing_newline(&buf);
272+
state = MBOX_AFTER_TRIPLE_DASH;
273+
}
274+
} else if (state == MBOX_IN_DIFF) {
275+
switch (line[0]) {
276+
case '\0': /* newer GNU diff, an empty context line */
277+
case '+':
278+
case '-':
279+
case ' ':
280+
/* A `-- ` line indicates the end of a diff */
281+
if (!old_count && !new_count)
282+
break;
283+
if (old_count && line[0] != '+')
284+
old_count--;
285+
if (new_count && line[0] != '-')
286+
new_count--;
287+
/* fallthrough */
288+
case '\\':
289+
strbuf_addstr(&buf, line);
290+
strbuf_addch(&buf, '\n');
291+
util->diffsize++;
292+
continue;
293+
case '@':
294+
if (parse_hunk_header(line, &old_count,
295+
&new_count, &p))
296+
break;
297+
298+
strbuf_addstr(&buf, "@@");
299+
if (current_filename && *p)
300+
strbuf_addf(&buf, " %s:",
301+
current_filename);
302+
strbuf_addstr(&buf, p);
303+
strbuf_addch(&buf, '\n');
304+
util->diffsize++;
305+
continue;
306+
default:
307+
if (old_count || new_count)
308+
warning(_("diff ended prematurely (-%d/+%d)"),
309+
(int)old_count, (int)new_count);
310+
break;
311+
}
312+
313+
if (util) {
314+
string_list_append(list, buf.buf)->util = util;
315+
util = NULL;
316+
strbuf_reset(&buf);
317+
}
318+
state = MBOX_BEFORE_HEADER;
319+
goto parse_from_delimiter;
320+
}
321+
}
322+
strbuf_release(&contents);
323+
324+
if (util) {
325+
if (state == MBOX_IN_DIFF)
326+
string_list_append(list, buf.buf)->util = util;
327+
else
328+
free(util);
329+
}
330+
strbuf_release(&buf);
331+
strbuf_release(&long_subject);
332+
free(current_filename);
333+
334+
return 0;
335+
}
336+
29337
/*
30338
* Reads the patches into a string list, with the `util` field being populated
31339
* as struct object_id (will need to be free()d).
@@ -41,6 +349,10 @@ static int read_patches(const char *range, struct string_list *list,
41349
ssize_t len;
42350
size_t size;
43351
int ret = -1;
352+
const char *path;
353+
354+
if (skip_prefix(range, "mbox:", &path))
355+
return read_mbox(path, list);
44356

45357
strvec_pushl(&cp.args, "log", "--no-color", "-p", "--no-merges",
46358
"--reverse", "--date-order", "--decorate=no",
@@ -424,6 +736,19 @@ static void output_pair_header(struct diff_options *diffopt,
424736

425737
strbuf_addch(buf, ' ');
426738
pp_commit_easy(CMIT_FMT_ONELINE, commit, buf);
739+
} else {
740+
struct patch_util *util = b_util ? b_util : a_util;
741+
const char *needle = "\n ## Commit message ##\n";
742+
const char *p = !util || !util->patch ?
743+
NULL : strstr(util->patch, needle);
744+
if (p) {
745+
if (status == '!')
746+
strbuf_addf(buf, "%s%s", color_reset, color);
747+
748+
strbuf_addch(buf, ' ');
749+
p += strlen(needle);
750+
strbuf_add(buf, p, strchrnul(p, '\n') - p);
751+
}
427752
}
428753
strbuf_addf(buf, "%s\n", color_reset);
429754

@@ -554,6 +879,9 @@ int show_range_diff(const char *range1, const char *range2,
554879
if (range_diff_opts->left_only && range_diff_opts->right_only)
555880
res = error(_("options '%s' and '%s' cannot be used together"), "--left-only", "--right-only");
556881

882+
if (!strcmp(range1, "mbox:-") && !strcmp(range2, "mbox:-"))
883+
res = error(_("only one mbox can be read from stdin"));
884+
557885
if (!res && read_patches(range1, &branch1, range_diff_opts->other_arg))
558886
res = error(_("could not parse log for '%s'"), range1);
559887
if (!res && read_patches(range2, &branch2, range_diff_opts->other_arg))
@@ -575,10 +903,18 @@ int show_range_diff(const char *range1, const char *range2,
575903
int is_range_diff_range(const char *arg)
576904
{
577905
char *copy = xstrdup(arg); /* setup_revisions() modifies it */
578-
const char *argv[] = { "", copy, "--", NULL };
906+
const char *argv[] = { "", copy, "--", NULL }, *path;
579907
int i, positive = 0, negative = 0;
580908
struct rev_info revs;
581909

910+
if (skip_prefix(arg, "mbox:", &path)) {
911+
free(copy);
912+
if (!strcmp(path, "-") || file_exists(path))
913+
return 1;
914+
error_errno(_("not an mbox: '%s'"), path);
915+
return 0;
916+
}
917+
582918
init_revisions(&revs, NULL);
583919
if (setup_revisions(3, argv, &revs, NULL) == 1) {
584920
for (i = 0; i < revs.pending.nr; i++)

0 commit comments

Comments
 (0)