CDRIVER-4119 Replace ICU SASLprep with internal implementation #1317

joshbsiegel · 2023-06-21T17:40:15Z

Summary

This PR will replace the ICU library that was used for SASLprep with an internal implementation of SASLprep. Once merged, a separate PR will be made to remove the ENABLE_ICU option and related code.

Background

SASLprep defines the steps required to prepare usernames and passwords for comparison and authentication. This is necessary as the vast amount of UTF-8 characters that a user can input for a password presents the need for some standardization. For the C driver, we use it to prepare passwords when authenticating with SCRAM-SHA-256. SASLprep is simply a profile built upon stringprep that explains specifically what parts of the stringprep document to abide by. From a high level, here is what is required by SASLprep:

Mapping: There are some characters that are similar to a space character and should be mapped to that. There are also characters that should be mapped to nothing (deletion).
Normalization: SASLprep requires Unicode normalization form KC. There are many changes that this normalization will make. For example, a character such as "Ⅳ" (one character) will be modified to "IV" (two characters).
Prohibited characters: Some characters are fully prohibited.
Bidirectional characters: While most characters are left to right, there are some characters that are right to left. These characters are allowed but must be checked for some requirements (see code comments).

Previously SASLprep was done with the ICU library and the goal of this ticket is to get rid of that dependency.

utf8proc

The normalization described above is the most involved step and we opted to use a library to handle it. utf8proc is bundled as a part of this ticket and handles this.

Preliminary Work

All of the references to characters in the stringprep and SASLprep RFC documents (such as what characters are mapped to what, prohibited characters, etc.) are displayed using their Unicode codepoints. This is not the same as a character's byte encoding in UTF-8 (see here). Because of this, all of the lookup tables that are required use Unicode codepoints, which is different than the const char *in_utf8 parameter in the existing _mongoc_sasl_prep_impl function. As such, it was necessary to convert the UTF-8 characters to their Unicode codepoints before doing any other work.

UTF-8 characters can be 1-4 bytes, so that is why _mongoc_utf8_string_length and _mongoc_utf8_char_length were added instead of using strlen (which will return the number of bytes in the string). The length of each UTF8 character can be determined by the first byte. _mongoc_utf8_to_unicode handles the conversion.

Lookup Tables

The mapping, prohibited characters, and bidirectional steps all use lookup tables to easily find whether a codepoint belongs in a particular category. The tables work by using ranges, where the 2n element is the lower range and the 2n + 1 element is the upper range (both inclusive). _mongoc_is_code_in_table will search through the table passed through and return whether the codepoint exists in the table. There might be a potential improvement here as I believe a binary search could be implemented for a faster lookup.

Misc

All of the tables and much of the instructions on how to perform SASLprep actually come from the stringprep RFC document. The SASLprep document just specifies what parts of stringprep to follow (which is most of it).

What's New

_mongoc_sasl_prep_impl was almost completely scrapped for the SASLprep steps listed above
_mongoc_utf8_char_length
_mongoc_utf8_string_length
_mongoc_utf8_is_valid
_mongoc_char_between_chars
_mongoc_is_code_in_table
_mongoc_utf8_to_unicode
_mongoc_unicode_to_utf8
_mongoc_unicode_codepoint_length
Tests added to test_mongoc_scram_sasl_prep
test_mongoc_utf8_char_length
test_mongoc_utf8_string_length
test_mongoc_utf8_to_unicode

joshbsiegel · 2023-06-21T17:46:42Z

src/libmongoc/src/mongoc/mongoc-scram-private.h

There were a few functions and tables that I needed to add to this file (and mongoc-scram.cpp) and I believe they belong elsewhere to improve organization but I wasn't sure where to put them as I wanted to follow the codebase structure/protocol. Please let me know what suggestions you have on where to place these additions and it should be an easy change.

Suggest moving the declarations only needed by mongoc-scram.c into mongoc-scram.c (e.g. _mongoc_utf8_is_valid). Declaring in the header may suggest they are intended to be used in other files.

Suggest adding a comment to symbols that are only included in the header for testing purposes (e.g. _mongoc_utf8_char_length ): is exposed for testing.

kevinAlbs

I suggest splitting the bundling of utf8proc into a separate PR. That may help keep the scope of the review smaller and to get passing on CI.

I have an open question about the bundling of utf8proc and I want input from others on the team. I can post this open question on the separate PR:

If utf8proc is bundled, is an option needed to use a system install of utf8proc?
- Would Debian and other package managers require having an option to use a system install of utf8proc if available?
- IMO utf8proc is a small library. I prefer not having another option between using a system install or the bundled if it is not needed.

CMakeLists.txt

src/utf8proc-2.8.0/LICENSE.md

build/cmake/make_dist/MakeDistCheck.cmake

src/libmongoc/CMakeLists.txt

src/CMakeLists.txt

kevinAlbs · 2023-06-22T15:54:47Z

src/libmongoc/CMakeLists.txt

+   COPYONLY
+)
+set (UTF8PROC_INCLUDE_DIRS "")
+# if (ENABLE_UTF8PROC MATCHES "SYSTEM|AUTO")


Remove commented out section.

src/libmongoc/CMakeLists.txt

src/libmongoc/src/mongoc/mongoc-scram-private.h

src/libmongoc/src/mongoc/mongoc-scram.c

src/libmongoc/tests/test-mongoc-scram.c

adriandole · 2023-07-10T17:23:47Z

src/libmongoc/src/mongoc/mongoc-scram-private.h

+int
+_mongoc_unicode_to_utf8 (uint32_t c, uint8_t *out);
+
+/* the tables below all fome from RFC 3454. They are all range tables, with


Suggested change

/* the tables below all fome from RFC 3454. They are all range tables, with

/* the tables below all come from RFC 3454. They are all range tables, with

eramongodb · 2023-07-11T16:50:59Z

src/libmongoc/src/mongoc/mongoc-scram-private.h

@@ -123,6 +123,348 @@ _mongoc_sasl_prep_required (const char *str);
 char *
 _mongoc_sasl_prep (const char *in_utf8, int in_utf8_len, bson_error_t *err);

+/* returns how many bytes a UTF-8 character is. */
+size_t
+_mongoc_utf8_char_length (const uint8_t *c);


The suggestion to use/cast-to uint8_t was intended to be within the implementation, not in the interface, of the utf8 functions. Keeping the interface char* will avoid requiring users to explicitly cast to uint8_t as currently being done in the updated test cases.

Suggest reverting the function declarations to use char* while still defining their operations in terms of uint8_t in their implementation.

eramongodb · 2023-07-11T16:55:30Z

src/libmongoc/src/mongoc/mongoc-scram.c

+      const size_t utf8_char_length = _mongoc_utf8_char_length (c);
+
+      if (!_mongoc_utf8_is_valid (c, utf8_char_length))
+         return -1;


If we really must return a special value to indicate invalid lengths (instead of requiring/asserting that the given UTF-8 string is valid and length computation will always succeed), may need to use ssize_t for the return type of this function to handle negative values.

eramongodb · 2023-07-11T16:57:06Z

src/libmongoc/src/mongoc/mongoc-scram-private.h

+
+/* converts a Unicode code point to UTF-8 character. Returns how many bytes the
+ * character converted is. */
+int


Either size_t (if we forbid invalid UTF-8 strings as input) or ssize_t (if we must return -1 on invalid input).

eramongodb · 2023-07-11T16:57:46Z

src/libmongoc/src/mongoc/mongoc-scram-private.h

+/* converts a Unicode code point to UTF-8 character. Returns how many bytes the
+ * character converted is. */
+int
+_mongoc_unicode_to_utf8 (uint32_t c, uint8_t *out);


Suggested change

_mongoc_unicode_to_utf8 (uint32_t c, uint8_t *out);

_mongoc_utf8_code_point_to_str (uint32_t c, uint8_t *out);

Suggest renaming to more accurate description + consistency with other UTF-8 functions.

eramongodb · 2023-07-11T17:00:57Z

src/libmongoc/src/mongoc/mongoc-scram.c

@@ -1035,12 +1036,12 @@ _mongoc_sasl_prep_impl (const char *name,
                        int in_utf8_len,


This parameter is now unused: should _mongoc_utf8_string_length account for the possibility of in_utf8_len < strlen(in_utf8)?

I'm a little unsure on what you mean by this. In all cases of _mongoc_sasl_prep_impl being called, in_utf8_len is passed through as strlen(in_utf8), meaning that in_utf8_len == strlen(in_utf8).

Good point. I suggest removing the redundant/unused parameter from _mongoc_sasl_prep and _mongoc_sasl_prep_impl to address the unused parameter warning(s).

eramongodb · 2023-07-11T17:04:06Z

src/libmongoc/src/mongoc/mongoc-scram-private.h

+ * is that the 2*n element is the lower bound and the 2*n + 1 is the upper bound
+ * (both inclusive). */
+bool
+_mongoc_is_code_point_in_table (uint32_t code,


Suggested change

_mongoc_is_code_point_in_table (uint32_t code,

_mongoc_utf8_code_point_is_in_table (uint32_t code,

Suggest renaming for consistency with other UTF-8 functions.

eramongodb · 2023-07-11T17:04:21Z

src/libmongoc/src/mongoc/mongoc-scram-private.h

+
+/* returns how many bytes a unicode codepoint is. */
+size_t
+_mongoc_unicode_codepoint_length (uint32_t c);


Suggested change

_mongoc_unicode_codepoint_length (uint32_t c);

_mongoc_utf8_code_point_length (uint32_t c);

Suggest renaming for consistency with other UTF-8 functions.

eramongodb · 2023-07-11T17:08:41Z

src/libmongoc/src/mongoc/mongoc-scram.c

+
+   uint8_t *loc = utf8_pre_norm;
+   for (size_t i = 0; i < num_chars; ++i) {
+      utf8_char_length = _mongoc_unicode_to_utf8 (utf8_codepoints[i], loc);


Missing validity check.

eramongodb · 2023-07-11T17:08:47Z

src/libmongoc/src/mongoc/mongoc-scram.c

+      utf8_pre_norm_len +=
+         _mongoc_unicode_codepoint_length (utf8_codepoints[i]);


Missing validity check.

src/libmongoc/src/mongoc/mongoc-scram.c

eramongodb

Minor suggestions; otherwise, LGTM.

eramongodb · 2023-07-11T20:03:36Z

src/libmongoc/src/mongoc/mongoc-scram.c

+
+   const uint8_t *c = (uint8_t *) s;
+
+   size_t str_length = 0;


Suggested change

size_t str_length = 0;

ssize_t str_length = 0;

Consistent signedness with return value.

eramongodb · 2023-07-11T20:04:37Z

src/libmongoc/tests/test-mongoc-scram.c

+static void
+test_mongoc_utf8_string_length (void)
+{
+   ASSERT_CMPINT (_mongoc_utf8_string_length (",ase"), ==, 4u);


Suggested change

ASSERT_CMPINT (_mongoc_utf8_string_length (",ase"), ==, 4u);

ASSERT_CMPSIZE_T (_mongoc_utf8_string_length (",ase"), ==, 4u);

Use ASSERT_CMPSIZE_T to compare size_t.

eramongodb · 2023-07-11T20:05:22Z

src/libmongoc/tests/test-mongoc-scram.c

+static void
+test_mongoc_utf8_to_unicode (void)
+{
+   ASSERT_CMPINT (_mongoc_utf8_get_first_code_point (",", 1), ==, 0x002C);


Suggested change

ASSERT_CMPINT (_mongoc_utf8_get_first_code_point (",", 1), ==, 0x002C);

ASSERT_CMPUINT32 (_mongoc_utf8_get_first_code_point (",", 1), ==, 0x002C);

Use ASSERT_CMPUINT32 to compare uint32_t.

eramongodb · 2023-07-11T20:18:25Z

src/libmongoc/src/mongoc/mongoc-scram-private.h

+size_t
+_mongoc_utf8_char_length (const char *c);
+
+/* returns how many characters are in a UTF-8 string. Returns -1 on error. */


Suggested change

/* returns how many characters are in a UTF-8 string. Returns -1 on error. */

/* Returns the byte length of the UTF-8 string. Returns -1 if `s` is not a valid

* UTF-8 string. */

eramongodb · 2023-07-11T20:18:59Z

src/libmongoc/src/mongoc/mongoc-scram-private.h

-_mongoc_sasl_prep (const char *in_utf8, int in_utf8_len, bson_error_t *err);
+_mongoc_sasl_prep (const char *in_utf8, bson_error_t *err);
+
+/* returns how many bytes a UTF-8 character is. */


Suggested change

/* returns how many bytes a UTF-8 character is. */

/* Returns the byte length of the first UTF-8 code point in `s`. */

eramongodb · 2023-07-11T20:20:16Z

src/libmongoc/src/mongoc/mongoc-scram-private.h

+ssize_t
+_mongoc_utf8_string_length (const char *s);
+
+/* returns whether a UTF-8 character is valid or not. */


Suggested change

/* returns whether a UTF-8 character is valid or not. */

/* Returns true if the first UTF-8 code point in `s` is valid. */

eramongodb · 2023-07-11T20:24:41Z

src/libmongoc/src/mongoc/mongoc-scram-private.h

+                                     const uint32_t *table,
+                                     size_t size);
+
+/* returns the first Unicode codepoint of a UTF-8 character. */


Suggested change

/* returns the first Unicode codepoint of a UTF-8 character. */

/* Returns the first Unicode code point in `c`. Returns 0 if length is 0.

* `c` must be a valid UTF-8 string. */

eramongodb · 2023-07-11T20:25:21Z

src/libmongoc/src/mongoc/mongoc-scram-private.h

+/* returns how many bytes a unicode codepoint is. Returns -1 on invalid code
+ * point. */


Suggested change

/* returns how many bytes a unicode codepoint is. Returns -1 on invalid code

* point. */

/* Returns the byte length of the UTF-8 code point. Returns -1 if `c` is not a

* valid UTF-8 code point. */

kevinAlbs

Great work. LGTM with minor comments addressed.

kevinAlbs · 2023-07-12T19:36:49Z

src/libmongoc/src/mongoc/mongoc-scram-private.h

+/* the tables below all come from RFC 3454. They are all range tables, with
+ * value 2*n being the lower range, and value 2*n + 1 being the upper range.
+ * Both ranges are exclusive.
+ */


Suggested change

/* the tables below all come from RFC 3454. They are all range tables, with

* value 2*n being the lower range, and value 2*n + 1 being the upper range.

* Both ranges are exclusive.

*/

/* the tables below all come from RFC 3454. They are all range tables, with

* value 2*n being the lower bound, and value 2*n + 1 being the upper bound.

* Both bounds are inclusive.

*/

Suggest using bound to refer to the range bounds. Comment above _mongoc_utf8_code_point_is_in_table and implementation suggests the bounds are inclusive.

kevinAlbs · 2023-07-13T16:00:19Z

src/libmongoc/src/mongoc/mongoc-scram-private.h

Suggest moving the declarations only needed by mongoc-scram.c into mongoc-scram.c (e.g. _mongoc_utf8_is_valid). Declaring in the header may suggest they are intended to be used in other files.

Suggest adding a comment to symbols that are only included in the header for testing purposes (e.g. _mongoc_utf8_char_length ): is exposed for testing.

kevinAlbs · 2023-07-13T17:05:18Z

src/libmongoc/src/mongoc/mongoc-scram-private.h

@@ -121,7 +121,354 @@ _mongoc_sasl_prep_required (const char *str);
 * null on error and sets err.
 * `name` should be "username" or "password". */


Suggested change

* `name` should be "username" or "password". */

kevinAlbs · 2023-07-13T17:05:36Z

src/libmongoc/src/mongoc/mongoc-scram-private.h

+
+/* returns the byte length of the first UTF-8 code point in `s`. */
+size_t
+_mongoc_utf8_char_length (const char *c);


Suggested change

_mongoc_utf8_char_length (const char *c);

_mongoc_utf8_char_length (const char *s);

kevinAlbs · 2023-07-13T17:07:04Z

src/libmongoc/src/mongoc/mongoc-scram-private.h

+
+/* returns true if the first UTF-8 code point in `s` is valid. */
+bool
+_mongoc_utf8_is_valid (const char *c, size_t length);


Suggested change

_mongoc_utf8_is_valid (const char *c, size_t length);

_mongoc_utf8_first_code_point_is_valid (const char *s, size_t length);

Suggest renaming to emphasize only the first code point is considered, not the full string.

kevinAlbs · 2023-07-13T17:21:11Z

src/libmongoc/src/mongoc/mongoc-scram.c

-   if (error_code) {
+   /* convert to unicode. */
+   utf8_codepoints = bson_malloc (sizeof (uint32_t) *
+                                  (num_chars + 1)); /* add one for null byte. */


Suggested change

(num_chars + 1)); /* add one for null byte. */

(num_chars + 1)); /* add one for trailing 0 value. */

Last value in utf8_codepoints is a 0 uint32_t, not one byte.

kevinAlbs · 2023-07-13T18:06:58Z

src/libmongoc/src/mongoc/mongoc-scram.c

 #ifdef MONGOC_ENABLE_ICU
-   return _mongoc_sasl_prep_impl ("password", in_utf8, in_utf8_len, err);
+   return _mongoc_sasl_prep_impl ("password", in_utf8, err);


Suggested change

return _mongoc_sasl_prep_impl ("password", in_utf8, err);

if (_mongoc_sasl_prep_required (in_utf8)) {

return _mongoc_sasl_prep_impl ("password", in_utf8, err);

}

return bson_strdup (in_utf8);

Suggest only running _mongoc_sasl_prep_impl if a password includes non-ASCII characters. I expect the majority of passwords are ASCII only characters and do not need to undergo SASLPrep.

kevinAlbs · 2023-07-13T18:11:07Z

src/libmongoc/src/mongoc/mongoc-scram.c

+   num_chars = curr;
+
+
+   // b. Normalize - normalize the result of step 1 using Unicode


Suggested change

// b. Normalize - normalize the result of step 1 using Unicode

// b. Normalize - normalize the result of step `a` using Unicode

kevinAlbs · 2023-07-13T18:18:12Z

src/libmongoc/src/mongoc/mongoc-scram.c

+         if (contains_RandALCar)
+            break;


Suggested change

if (contains_RandALCar)

break;

Suggest removing unnecessary early break for consistency with assignment of contains_RandALCar = true;

joshbsiegel added 11 commits June 14, 2023 19:27

add UTF8 string helper functions

e494861

utf8 to unicode and tests

8bc5714

mapping stage, tables, and table related funcitons

60d4df9

prohibit and bidi steps

90fedac

find output length

7a8e3fa

saslprep completion minus normalization, tests

5b67586

bundle utf8proc

1acb090

add normalization

56a0d54

another normalization test

020c0fa

change variable names for clarity

9779387

minor typos

f3e443e

joshbsiegel commented Jun 21, 2023

View reviewed changes

joshbsiegel requested a review from kevinAlbs June 21, 2023 18:01

fixing comments

71b8c44

kevinAlbs reviewed Jun 22, 2023

View reviewed changes

joshbsiegel mentioned this pull request Jun 22, 2023

CDRIVER-4668 Add utf8proc library and update CMake configuration #1319

Merged

joshbsiegel changed the title ~~CDRIVER-4119 Replace ICU SASLPrep with internal implementation~~ CDRIVER-4119 Replace ICU SASLprep with internal implementation Jun 27, 2023

joshbsiegel added 5 commits July 7, 2023 15:55

Merge branch 'master' into cdriver-4119

01dbaa2

removing utf8proc files

875a733

sync cmake config with master

1f548fb

eof endline

a3f674e

update utf8proc.h include

bd1d185

joshbsiegel requested review from vector-of-bool, adriandole and eramongodb July 7, 2023 17:51

eramongodb requested changes Jul 7, 2023

View reviewed changes

joshbsiegel added 4 commits July 10, 2023 15:20

change types to uint32_t and uint8_t, assert parameters

5a0bfc1

moving function impl out of ifdef

6572af3

moving tests into enable_ssl for evergreen

d940cfa

modify to <utf8proc.h>

92ebf33

adriandole suggested changes Jul 10, 2023

View reviewed changes

joshbsiegel added 3 commits July 10, 2023 18:03

add utf8proc_include_dirs

7e624b1

eof newline

31d2c83

free out_utf8 on error

bf53c21

joshbsiegel requested review from adriandole and eramongodb July 10, 2023 20:20

eramongodb reviewed Jul 11, 2023

View reviewed changes

joshbsiegel added 2 commits July 11, 2023 18:47

revert definitions to char *, change function names, validity checks

40d6ae3

remove unused length parameter

6d0b012

joshbsiegel requested a review from eramongodb July 11, 2023 19:52

eramongodb approved these changes Jul 11, 2023

View reviewed changes

comment fixes and test asserts

e3f1760

joshbsiegel requested a review from kevinAlbs July 11, 2023 20:36

adriandole approved these changes Jul 13, 2023

View reviewed changes

joshbsiegel removed the request for review from vector-of-bool July 13, 2023 18:29

kevinAlbs approved these changes Jul 13, 2023

View reviewed changes

relocate function definitions and fix comments

8c8c426

joshbsiegel merged commit 4b9da57 into mongodb:master Jul 13, 2023

joshbsiegel mentioned this pull request Jul 15, 2023

CDRIVER-4680 Remove ENABLE_ICU option and related code #1342

Merged

GreenReaper mentioned this pull request Dec 19, 2023

Support SCRAM-SHA-256 auth mechanism EnterpriseDB/mongo_fdw#174

Closed

	/* the tables below all fome from RFC 3454. They are all range tables, with
	/* the tables below all come from RFC 3454. They are all range tables, with

	_mongoc_unicode_to_utf8 (uint32_t c, uint8_t *out);
	_mongoc_utf8_code_point_to_str (uint32_t c, uint8_t *out);

		@@ -1035,12 +1036,12 @@ _mongoc_sasl_prep_impl (const char *name,
		int in_utf8_len,

	_mongoc_is_code_point_in_table (uint32_t code,
	_mongoc_utf8_code_point_is_in_table (uint32_t code,

	_mongoc_unicode_codepoint_length (uint32_t c);
	_mongoc_utf8_code_point_length (uint32_t c);

		utf8_pre_norm_len +=
		_mongoc_unicode_codepoint_length (utf8_codepoints[i]);

	ASSERT_CMPINT (_mongoc_utf8_string_length (",ase"), ==, 4u);
	ASSERT_CMPSIZE_T (_mongoc_utf8_string_length (",ase"), ==, 4u);

	ASSERT_CMPINT (_mongoc_utf8_get_first_code_point (",", 1), ==, 0x002C);
	ASSERT_CMPUINT32 (_mongoc_utf8_get_first_code_point (",", 1), ==, 0x002C);

	/* returns how many characters are in a UTF-8 string. Returns -1 on error. */
	/* Returns the byte length of the UTF-8 string. Returns -1 if `s` is not a valid
	* UTF-8 string. */

	/* returns how many bytes a UTF-8 character is. */
	/* Returns the byte length of the first UTF-8 code point in `s`. */

	/* returns whether a UTF-8 character is valid or not. */
	/* Returns true if the first UTF-8 code point in `s` is valid. */

	/* returns the first Unicode codepoint of a UTF-8 character. */
	/* Returns the first Unicode code point in `c`. Returns 0 if length is 0.
	* `c` must be a valid UTF-8 string. */

		/* returns how many bytes a unicode codepoint is. Returns -1 on invalid code
		* point. */

		@@ -121,7 +121,354 @@ _mongoc_sasl_prep_required (const char *str);
		* null on error and sets err.
		* `name` should be "username" or "password". */

	_mongoc_utf8_char_length (const char *c);
	_mongoc_utf8_char_length (const char *s);

	_mongoc_utf8_is_valid (const char *c, size_t length);
	_mongoc_utf8_first_code_point_is_valid (const char *s, size_t length);

	(num_chars + 1)); /* add one for null byte. */
	(num_chars + 1)); /* add one for trailing 0 value. */

		num_chars = curr;


		// b. Normalize - normalize the result of step 1 using Unicode

CDRIVER-4119 Replace ICU SASLprep with internal implementation #1317

CDRIVER-4119 Replace ICU SASLprep with internal implementation #1317

Uh oh!

Conversation

joshbsiegel commented Jun 21, 2023 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Summary

Background

Preliminary Work

Lookup Tables

Misc

What's New

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

kevinAlbs left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

eramongodb left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

joshbsiegel commented Jun 21, 2023 •

edited

Loading