builtin: patch-id: add --verbatim as a command mode
There are situations where the user might not want the default setting where patch-id strips all whitespace. They might be working in a language where white space is syntactically important, or they might have CI testing that enforces strict whitespace linting. In these cases, a whitespace change would result in the patch fundamentally changing, and thus deserving of a different id. Add a new mode that is exclusive of --stable and --unstable called --verbatim. It also corresponds to the config patchid.verbatim = true. In this mode, the stable algorithm is used and whitespace is not stripped from the patch text. Users of --unstable mainly care about compatibility with old git versions, which unstripping the whitespace would break. Thus there isn't a usecase for the combination of --verbatim and --unstable, and we don't expose this so as to not add maintainence burden. Signed-off-by: Jerry Zhang <jerry@skydio.com> fixes https://github.com/Skydio/revup/issues/2 Signed-off-by: Junio C Hamano <gitster@pobox.com>
Jerry Zhang committed
Oct 24, 2022 at 20:07 UTC
2871f4d447214874e13cf764ab3a170c9d844ca2
3 files changed
+124
-39
Documentation/git-patch-id.txt
+16
-8
@@ -8,18 +8,18 @@ git-patch-id - Compute unique ID for a patch
8
SYNOPSIS
9
--------
10
[verse]
11
-'git patch-id' [--stable | --unstable]
11
+'git patch-id' [--stable | --unstable | --verbatim]
12
13
DESCRIPTION
14
-----------
15
Read a patch from the standard input and compute the patch ID for it.
16
17
A "patch ID" is nothing but a sum of SHA-1 of the file diffs associated with a
18
-patch, with whitespace and line numbers ignored. As such, it's "reasonably
19
-stable", but at the same time also reasonably unique, i.e., two patches that
20
-have the same "patch ID" are almost guaranteed to be the same thing.
18
+patch, with line numbers ignored. As such, it's "reasonably stable", but at
19
+the same time also reasonably unique, i.e., two patches that have the same
20
+"patch ID" are almost guaranteed to be the same thing.
21
22
-IOW, you can use this thing to look for likely duplicate commits.
22
+The main usecase for this command is to look for likely duplicate commits.
23
24
When dealing with 'git diff-tree' output, it takes advantage of
25
the fact that the patch is prefixed with the object name of the
@@ -30,6 +30,12 @@ This can be used to make a mapping from patch ID to commit ID.
30
OPTIONS
31
-------
32
33
+--verbatim::
34
+ Calculate the patch-id of the input as it is given, do not strip
35
+ any whitespace.
36
+
37
+ This is the default if patchid.verbatim is true.
38
+
39
--stable::
40
Use a "stable" sum of hashes as the patch ID. With this option:
41
- Reordering file diffs that make up a patch does not affect the ID.
@@ -45,14 +51,16 @@ OPTIONS
51
of "-O<orderfile>", thereby making existing databases storing such
52
"unstable" or historical patch-ids unusable.
53
54
+ - All whitespace within the patch is ignored and does not affect the id.
55
+
56
This is the default if patchid.stable is set to true.
57
58
--unstable::
59
Use an "unstable" hash as the patch ID. With this option,
60
the result produced is compatible with the patch-id value produced
53
- by git 1.9 and older. Users with pre-existing databases storing
54
- patch-ids produced by git 1.9 and older (who do not deal with reordered
55
- patches) may want to use this option.
61
+ by git 1.9 and older and whitespace is ignored. Users with pre-existing
62
+ databases storing patch-ids produced by git 1.9 and older (who do not deal
63
+ with reordered patches) may want to use this option.
64
65
This is the default.
66
builtin/patch-id.c
+48
-25
@@ -2,6 +2,7 @@
2
#include "builtin.h"
3
#include "config.h"
4
#include "diff.h"
5
+#include "parse-options.h"
6
7
static void flush_current_id(int patchlen, struct object_id *id, struct object_id *result)
8
{
@@ -57,7 +58,7 @@ static int scan_hunk_header(const char *p, int *p_before, int *p_after)
58
}
59
60
static int get_one_patchid(struct object_id *next_oid, struct object_id *result,
60
- struct strbuf *line_buf, int stable)
61
+ struct strbuf *line_buf, int stable, int verbatim)
62
{
63
int patchlen = 0, found_next = 0;
64
int before = -1, after = -1;
@@ -76,8 +77,11 @@ static int get_one_patchid(struct object_id *next_oid, struct object_id *result,
77
if (!skip_prefix(line, "diff-tree ", &p) &&
78
!skip_prefix(line, "commit ", &p) &&
79
!skip_prefix(line, "From ", &p) &&
79
- starts_with(line, "\\ ") && 12 < strlen(line))
80
+ starts_with(line, "\\ ") && 12 < strlen(line)) {
81
+ if (verbatim)
82
+ the_hash_algo->update_fn(&ctx, line, strlen(line));
83
continue;
84
+ }
85
86
if (!get_oid_hex(p, next_oid)) {
87
found_next = 1;
@@ -152,8 +156,8 @@ static int get_one_patchid(struct object_id *next_oid, struct object_id *result,
156
if (line[0] == '+' || line[0] == ' ')
157
after--;
158
155
- /* Compute the sha without whitespace */
156
- len = remove_space(line);
159
+ /* Add line to hash algo (possibly removing whitespace) */
160
+ len = verbatim ? strlen(line) : remove_space(line);
161
patchlen += len;
162
the_hash_algo->update_fn(&ctx, line, len);
163
}
@@ -166,7 +170,7 @@ static int get_one_patchid(struct object_id *next_oid, struct object_id *result,
170
return patchlen;
171
}
172
169
-static void generate_id_list(int stable)
173
+static void generate_id_list(int stable, int verbatim)
174
{
175
struct object_id oid, n, result;
176
int patchlen;
@@ -174,21 +178,32 @@ static void generate_id_list(int stable)
178
179
oidclr(&oid);
180
while (!feof(stdin)) {
177
- patchlen = get_one_patchid(&n, &result, &line_buf, stable);
181
+ patchlen = get_one_patchid(&n, &result, &line_buf, stable, verbatim);
182
flush_current_id(patchlen, &oid, &result);
183
oidcpy(&oid, &n);
184
}
185
strbuf_release(&line_buf);
186
}
187
184
-static const char patch_id_usage[] = "git patch-id [--stable | --unstable]";
188
+static const char *const patch_id_usage[] = {
189
+ N_("git patch-id [--stable | --unstable | --verbatim]"), NULL
190
+};
191
+
192
+struct patch_id_opts {
193
+ int stable;
194
+ int verbatim;
195
+};
196
197
static int git_patch_id_config(const char *var, const char *value, void *cb)
198
{
188
- int *stable = cb;
199
+ struct patch_id_opts *opts = cb;
200
201
if (!strcmp(var, "patchid.stable")) {
191
- *stable = git_config_bool(var, value);
202
+ opts->stable = git_config_bool(var, value);
203
+ return 0;
204
+ }
205
+ if (!strcmp(var, "patchid.verbatim")) {
206
+ opts->verbatim = git_config_bool(var, value);
207
return 0;
208
}
209
@@ -197,21 +212,29 @@ static int git_patch_id_config(const char *var, const char *value, void *cb)
212
213
int cmd_patch_id(int argc, const char **argv, const char *prefix)
214
{
200
- int stable = -1;
201
-
202
- git_config(git_patch_id_config, &stable);
203
-
204
- /* If nothing is set, default to unstable. */
205
- if (stable < 0)
206
- stable = 0;
207
-
208
- if (argc == 2 && !strcmp(argv[1], "--stable"))
209
- stable = 1;
210
- else if (argc == 2 && !strcmp(argv[1], "--unstable"))
211
- stable = 0;
212
- else if (argc != 1)
213
- usage(patch_id_usage);
214
-
215
- generate_id_list(stable);
215
+ /* if nothing is set, default to unstable */
216
+ struct patch_id_opts config = {0, 0};
217
+ int opts = 0;
218
+ struct option builtin_patch_id_options[] = {
219
+ OPT_CMDMODE(0, "unstable", &opts,
220
+ N_("use the unstable patch-id algorithm"), 1),
221
+ OPT_CMDMODE(0, "stable", &opts,
222
+ N_("use the stable patch-id algorithm"), 2),
223
+ OPT_CMDMODE(0, "verbatim", &opts,
224
+ N_("don't strip whitespace from the patch"), 3),
225
+ OPT_END()
226
+ };
227
+
228
+ git_config(git_patch_id_config, &config);
229
+
230
+ /* verbatim implies stable */
231
+ if (config.verbatim)
232
+ config.stable = 1;
233
+
234
+ argc = parse_options(argc, argv, prefix, builtin_patch_id_options,
235
+ patch_id_usage, 0);
236
+
237
+ generate_id_list(opts ? opts > 1 : config.stable,
238
+ opts ? opts == 3 : config.verbatim);
239
return 0;
240
}
t/t4204-patch-id.sh
+60
-6
@@ -8,13 +8,13 @@ export GIT_TEST_DEFAULT_INITIAL_BRANCH_NAME
8
. ./test-lib.sh
9
10
test_expect_success 'setup' '
11
- as="a a a a a a a a" && # eight a
12
- test_write_lines $as >foo &&
13
- test_write_lines $as >bar &&
11
+ str="ab cd ef gh ij kl mn op" &&
12
+ test_write_lines $str >foo &&
13
+ test_write_lines $str >bar &&
14
git add foo bar &&
15
git commit -a -m initial &&
16
- test_write_lines $as b >foo &&
17
- test_write_lines $as b >bar &&
16
+ test_write_lines $str b >foo &&
17
+ test_write_lines $str b >bar &&
18
git commit -a -m first &&
19
git checkout -b same main &&
20
git commit --amend -m same-msg &&
@@ -22,8 +22,23 @@ test_expect_success 'setup' '
22
echo c >foo &&
23
echo c >bar &&
24
git commit --amend -a -m notsame-msg &&
25
+ git checkout -b with_space main~ &&
26
+ cat >foo <<-\EOF &&
27
+ a b
28
+ c d
29
+ e f
30
+ g h
31
+ i j
32
+ k l
33
+ m n
34
+ op
35
+ EOF
36
+ cp foo bar &&
37
+ git add foo bar &&
38
+ git commit --amend -m "with spaces" &&
39
test_write_lines bar foo >bar-then-foo &&
40
test_write_lines foo bar >foo-then-bar
41
+
42
'
43
44
test_expect_success 'patch-id output is well-formed' '
@@ -128,9 +143,21 @@ test_patch_id_file_order () {
143
git format-patch -1 --stdout -O foo-then-bar >format-patch.output &&
144
calc_patch_id <format-patch.output "ordered-$name" "$@" &&
145
cmp_patch_id $relevant "$name" "ordered-$name"
146
+}
147
148
+test_patch_id_whitespace () {
149
+ relevant="$1"
150
+ shift
151
+ name="ws-${1}-$relevant"
152
+ shift
153
+ get_top_diff "main~" >top-diff.output &&
154
+ calc_patch_id <top-diff.output "$name" "$@" &&
155
+ get_top_diff "with_space" >top-diff.output &&
156
+ calc_patch_id <top-diff.output "ws-$name" "$@" &&
157
+ cmp_patch_id $relevant "$name" "ws-$name"
158
}
159
160
+
161
# combined test for options: add more tests here to make them
162
# run with all options
163
test_patch_id () {
@@ -146,6 +173,14 @@ test_expect_success 'file order is relevant with --unstable' '
173
test_patch_id_file_order relevant --unstable --unstable
174
'
175
176
+test_expect_success 'whitespace is relevant with --verbatim' '
177
+ test_patch_id_whitespace relevant --verbatim --verbatim
178
+'
179
+
180
+test_expect_success 'whitespace is irrelevant without --verbatim' '
181
+ test_patch_id_whitespace irrelevant --stable --stable
182
+'
183
+
184
#Now test various option combinations.
185
test_expect_success 'default is unstable' '
186
test_patch_id relevant default
@@ -161,6 +196,17 @@ test_expect_success 'patchid.stable = false is unstable' '
196
test_patch_id relevant patchid.stable=false
197
'
198
199
+test_expect_success 'patchid.verbatim = true is correct and stable' '
200
+ test_config patchid.verbatim true &&
201
+ test_patch_id_whitespace relevant patchid.verbatim=true &&
202
+ test_patch_id irrelevant patchid.verbatim=true
203
+'
204
+
205
+test_expect_success 'patchid.verbatim = false is unstable' '
206
+ test_config patchid.verbatim false &&
207
+ test_patch_id relevant patchid.verbatim=false
208
+'
209
+
210
test_expect_success '--unstable overrides patchid.stable = true' '
211
test_config patchid.stable true &&
212
test_patch_id relevant patchid.stable=true--unstable --unstable
@@ -171,6 +217,11 @@ test_expect_success '--stable overrides patchid.stable = false' '
217
test_patch_id irrelevant patchid.stable=false--stable --stable
218
'
219
220
+test_expect_success '--verbatim overrides patchid.stable = false' '
221
+ test_config patchid.stable false &&
222
+ test_patch_id_whitespace relevant stable=false--verbatim --verbatim
223
+'
224
+
225
test_expect_success 'patch-id supports git-format-patch MIME output' '
226
get_patch_id main &&
227
git checkout same &&
@@ -225,7 +276,10 @@ test_expect_success 'patch-id handles no-nl-at-eof markers' '
276
EOF
277
calc_patch_id nonl <nonl &&
278
calc_patch_id withnl <withnl &&
228
- test_cmp patch-id_nonl patch-id_withnl
279
+ test_cmp patch-id_nonl patch-id_withnl &&
280
+ calc_patch_id nonl-inc-ws --verbatim <nonl &&
281
+ calc_patch_id withnl-inc-ws --verbatim <withnl &&
282
+ ! test_cmp patch-id_nonl-inc-ws patch-id_withnl-inc-ws
283
'
284
285
test_expect_success 'patch-id handles diffs with one line of before/after' '