Java transitive search #130

tmihalac · 2025-10-05T14:46:02Z

No description provided.

zvigrinberg

@tmihalac
Great Job!, It looks quite good, although there are parts that are not that readable.... ( especially in the java segmenter of methods and java language parser, but for the fairness i'd say it's extermely detailed and covering a lot of cases and edge cases)
Please add 2-4 UT tests functions to protect your logic from regressions, re-use the following 2 functions that i've created for the sake of consistency in checking ( it's UT tests and semi Integration test - checking integration of a given ecosystem with all the module of transitive code search tool

vulnerability-analysis/src/vuln_analysis/tools/test_transitive_code_search.py

Lines 100 to 122 in 3481519

    
           def set_input_for_next_run(git_repository: str, git_ref: str, included_extensions: list[str], 
        
                                      excluded_extensions: list[str]): 
        
               source_code_info = [SourceDocumentsInfo(type='code', git_repo=git_repository, 
        
                                                       ref=git_ref, include=included_extensions, 
        
                                                       exclude=excluded_extensions)] 
        
               sbom_info_input = ManualSBOMInfoInput(packages=[SBOMPackage(name="a", version="1.0", system="blabla")]) 
        
               morpheus_input = AgentMorpheusInput(image=ImageInfoInput(source_info=source_code_info, 
        
                                                                        sbom_info=sbom_info_input, 
        
                                                                        analysis_type=AnalysisType.IMAGE), 
        
                                                   scan=ScanInfoInput(vulns=[VulnInfo(vuln_id="CVE-2025-1234")])) 
        
               engine_input = AgentMorpheusEngineInput(input=morpheus_input, info=AgentMorpheusInfo()) 
        
               state: AgentMorpheusEngineState = AgentMorpheusEngineState(original_input=engine_input, 
        
                                                                          code_vdb_path="", 
        
                                                                          doc_vdb_path="", 
        
                                                                          code_index_path="", 
        
                                                                          cve_intel=[]) 
        
               ctx_state.set(state) 
        
           async def get_transitive_code_runner_function(): 
        
               transitive_code_search = transitive_search(config=TransitiveCodeSearchToolConfig(), builder=None) 
        
               async for function in transitive_code_search.gen: 
        
                   return function.single_fn

And this is an example of go UT that uses these 2 functions:

vulnerability-analysis/src/vuln_analysis/tools/test_transitive_code_search.py

Lines 59 to 69 in 3481519

    
           async def test_transitive_search_golang_4(): 
        
               transitive_code_search_runner_coroutine = await get_transitive_code_runner_function() 
        
               set_input_for_next_run(git_repository="https://github.com/openshift/oc-mirror", 
        
                                      git_ref="b137a53a5360a41a70432ea2bfc98a6cee6f7a4a", 
        
                                      included_extensions=["**/*.go"], 
        
                                      excluded_extensions=["go.mod", "go.sum", "**/*test*.go"]) 
        
               result = await transitive_code_search_runner_coroutine("github.com/mholt/archiver,Unarchive") 
        
               (path_found, list_path) = result 
        
               assert path_found is True 
        
               assert len(list_path) == 2

Please prepare 2-4 test functions with inputs related to java.
In addition, Please see my comments.

Thanks!.

src/vuln_analysis/utils/functions_parsers/golang_functions_parsers.py

zvigrinberg · 2025-10-15T10:10:19Z

src/vuln_analysis/utils/java_segmenters_with_methods.py

+        # Extract
+        methods = _extract_methods_anywhere(java_source, 0, len(java_source))
+        # Lambdas in a separate pass across the whole file (so we don't double-count)
+        lambdas = _extract_lambdas_in_range(java_source, 0, len(java_source))


@tmihalac Eventually you included lambdas in the feature? ( it seems that the segmenter parse them out from sources into documents)

The lambdas need more work, currently the parameter types are not derived yet and its not handled yet in the search logic I've opened a new jira for it https://issues.redhat.com/browse/APPENG-3848

zvigrinberg · 2025-10-15T10:31:48Z

src/vuln_analysis/utils/functions_parsers/java_functions_parsers.py

@@ -0,0 +1,3671 @@
+import os


@tmihalac This file is an enormous one and is very hard to maintain. ( not likely to be easily manageable) - almost 4k lines of code.
i think, that all functions that are used internally and methods( should be transformed into functions before that), that are not part of the LanguageParser API should be moved to java_utils module (you already created this file module).

I Would even split all of the util functions between two utils files modules (java_utils and another one)
Maybe according to function' type/kind/how much the code of it is "low level"

I agree its pretty overwhelming but even splitting it to 2 files will still give 2 files of roughly 2K in addition its not so simple as some of the internal methods relay on other public parser methods which causes cyclic dependencies

Yes but i barely can write comments next to lines of code in this file because it's so big ( it stuck and takes a lot of time just to mark the portion for the comment), so if i can't even effectively review it and write comments, then it's a problem.
Even more severe - Maintainability is very impacted, and also readability - It's very hard to troubleshoot a potential bug with such a huge file, even with debugger.
try to split it to another 2-3 files, plus this file itself, then it would keep each one of them more manageable and approximately ~1K lines of code each.
About closures ( internal functions) , you should decouple them from the outer methods if possible , but if not possible, because they're using the outer methods, move them together with the outer scope ( the method) outside ( if they're using the outer scope vars), after turning the outer method to function.
the closures ( internal functions) of a method in the class cannot access other methods in that class ( unless it's a static method or class method, which is another story, or that you explicitly expose the instance in the method' scope as a var), they can just access the scope of the containing method and other closures in the same depth ( siblings) and that's it, so an internal/private method that is not part of the LanguageParser API, can be transformed to a function, and moved/migrated to another file , together with all of its closures.

zvigrinberg · 2025-10-15T14:16:49Z

src/vuln_analysis/utils/java_chain_of_calls_retriever.py

+        if not exist:
+            self.documents_of_functions = [doc for doc in tqdm(self.documents, total=len(self.documents), desc="Filtering documents, leaving functions/methods ...")
+                              if self.language_parser.is_function(doc)]
+            save_to_cache(PICKLE_CACHE_DIRECTORY, code_source_info.git_repo, code_source_info.ref, self.documents_of_functions ,"documents_of_functions")
+        else:
+            self.documents_of_functions = documents_of_functions
+
+        documents_of_types, exist =  retrieve_from_cache(PICKLE_CACHE_DIRECTORY, code_source_info.git_repo, code_source_info.ref, "documents_of_types")
+        # filter out full code documents and functions/methods docs, retaining only types/classes docs
+        if not exist:
+            self.documents_of_types = [doc for doc in tqdm(self.documents, total=len(self.documents) , desc="Filtering documents, leaving types/classes ...")
+                                       if self.language_parser.is_type(doc)]
+            save_to_cache(PICKLE_CACHE_DIRECTORY, code_source_info.git_repo, code_source_info.ref, self.documents_of_types ,"documents_of_types")
+        else:
+            self.documents_of_types = documents_of_types


@tmihalac Why are you calling save_to_cache twice?
Why is the separation of the types/classes and methods is needed here?

If i understood correctly, based on the existing logic, you will end up with 3 pickles files, one for all document, one of the java classes and one for the java methods.

But you're anyway getting these from the unified list of documents ( classes + methods).
Is it saving you a lot of time ( does the list comprehension is a costy/time expensive in your test cases?

I will end up with a few pickle files, the list comprehensions all together take around 5 - 10 min, each time but its enough for them to be done once and reused each time thats why I save them to cache as seperate files

list comprehension takes 5-10 minutes for each kind?
How many documents are in the list can you remind me?

No, it takes all of them together 5-10 min and its being run multiple time by the transitive search tool

zvigrinberg · 2025-10-15T14:23:40Z

src/vuln_analysis/utils/functions_parsers/java_functions_parsers.py

+import os
+import re
+from dataclasses import dataclass
+from typing import Dict, Tuple, Any, List, Optional
+from tqdm import tqdm
+
+from langchain_core.documents import Document
+
+from .lang_functions_parsers import LanguageFunctionsParser
+from ..java_utils import extract_jar_name


@tmihalac You should add agent logger and use it across the implementation wherever applicable.

Didn't see the need of it

With such a huge file ( even after refactoring and moving out most of the logic, it will remain 500-1k lines of code) i would suggest adding such an agent logger and add some logs across the main logic of it and wherever applicable.

zvigrinberg · 2025-10-15T14:27:44Z

src/vuln_analysis/utils/functions_parsers/lang_functions_parsers.py

+    def is_package_imported(self, code_content: str, identifier: str, callee_package: str = "") -> bool:
+        pass
+
+    def extract_method_name_with_params(self, java_src: str) -> str | None:


@tmihalac Parameter name should be independent of ecosystem

Suggested change

def extract_method_name_with_params(self, java_src: str) -> str | None:

def extract_method_name_with_params(self, src: str) -> str | None:

Your right, missed it in refactor

zvigrinberg · 2025-10-15T14:28:38Z

src/vuln_analysis/utils/functions_parsers/golang_functions_parsers.py

+                        return True
+        return False
+
+    def extract_method_name_with_params(self, java_src: str) -> str | None:


Suggested change

def extract_method_name_with_params(self, java_src: str) -> str | None:

def extract_method_name_with_params(self, golang_src: str) -> str | None:

Your right, missed it in refactor

zvigrinberg · 2025-10-16T06:11:14Z

src/vuln_analysis/tools/test_transitive_code_search.py

+@pytest.mark.asyncio
+async def test_transitive_search_java_1():
+    transitive_code_search_runner_coroutine = await get_transitive_code_runner_function()
+    set_input_for_next_run(git_repository="https://github.com/cryostatio/cryostat",
+                           git_ref="8f753753379e9381429b476aacbf6890ef101438",
+                           included_extensions=["**/*.java"],
+                           excluded_extensions=["target/**/*",
+                                                "build/**/*",
+                                                "*.class",
+                                                ".gradle/**/*",
+                                                ".mvn/**/*",
+                                                ".gitignore",
+                                                "test/**/*",
+                                                "tests/**/*",
+                                                "src/test/**/*",
+                                                "pom.xml",
+                                                "build.gradle"])
+    result = await transitive_code_search_runner_coroutine("commons-beanutils:commons-beanutils:1.9.4,org.apache.commons.beanutils.PropertyUtilsBean.getProperty")
+    (path_found, list_path) = result
+    print(result)
+    assert path_found is False
+    assert len(list_path) is 1
+


@tmihalac Great!
If you are having more, try to add at least more 1-2 inputs to protect your logic from future changes.

The problem is I don't have more examples, Ronit is working on trying to provide more

@tmihalac You don't have to wait for him...
Use other repos, and check for methods that are actually being called, and assert for the path length and for the path_found expression ( the second time), assert the boolean result against true
It Doesn't necessarily has to be vulnerable methods, you're only checking your logic of chain of calls
You can even give the same repo with other reachable method of an artifact ( either transitive or direct one).

tmihalac · 2025-10-30T17:01:20Z

/retest

tmihalac · 2025-10-30T17:03:45Z

/retest

zvigrinberg · 2025-10-30T17:48:34Z

/retest

zvigrinberg

@tmihalac Thanks for your good and hard work.
Please see my comments, we need couple of changes ( nothing critical though).
also, after all changes, please run flows in all ecosystems end to end ( we need to do it until Integration tests implementation WIP is done).

zvigrinberg · 2025-11-03T08:12:02Z

src/vuln_analysis/tools/transitive_code_search.py

+    coc_retriever = get_chain_of_calls_retriever(ecosystem,
+                                                 documents,
+                                                 git_repo,
+                                                 query,
+                                                 code_source_info)


@tmihalac Please use kw to pass the args instead of positional args, because it's very confusing that way.

zvigrinberg · 2025-11-03T09:13:19Z

src/vuln_analysis/tools/transitive_code_search.py

-        ecosystem = file.read()
-    ecosystem = Ecosystem[ecosystem]
-    coc_retriever = ChainOfCallsRetriever(documents=documents, ecosystem=ecosystem, manifest_path=git_repo)
+    ecosystem = TransitiveCodeSearcher.get_ecosystem(git_repo)


@tmihalac The Problem with this is returning the ecosystem even if the download dependencies' install dependencies method is failing ( or getting the tree builder for the "found ecosystem"), and thus in such case it can continue all the way to do the transitive search even without 3rd party libs source codes , which is wrong.
in current logic, it's only save it to the file in case it completes both of the getting the tree builder, and installing the dependencies...
Please revert

zvigrinberg · 2025-11-03T09:14:39Z

src/vuln_analysis/utils/transitive_code_searcher_tool.py

+        ecosystem: Ecosystem = TransitiveCodeSearcher.get_ecosystem(git_repo_path)
+        if ecosystem is None:
+            return False


@tmihalac The Problem with this is returning the ecosystem even if the download dependencies' install dependencies method is failing ( or getting the tree builder for the "found ecosystem"), and thus in such case it can continue all the way to do the transitive search even without 3rd party libs source codes , which is wrong.
in current logic, it's only save it to the file in case it completes both of the getting the tree builder, and installing the dependencies...
Please revert

zvigrinberg · 2025-11-03T09:19:37Z

src/vuln_analysis/utils/functions_parsers/python_functions_parser.py

+    def is_doc_type(self, doc: Document) -> bool:
+        return doc.page_content.startswith("type")
+


@tmihalac This is incorrect, the type in python is class
Why did you change it ?

vulnerability-analysis/src/vuln_analysis/utils/functions_parsers/python_functions_parser.py

Lines 99 to 100 in 0197565

def get_type_reserved_word(self) -> str:

return "class"

@tmihalac In addition, you should also instead method

Suggested change

def is_doc_type(self, doc: Document) -> bool:

return doc.page_content.startswith("type")

def is_doc_type(self, doc: Document) -> bool:

return doc.page_content.startswith(self.get_type_reserved_word())

You are right

zvigrinberg · 2025-11-03T09:29:19Z

src/vuln_analysis/utils/functions_parsers/python_functions_parser.py

+    def is_package_imported(self, code_content: str, identifier: str, callee_package: str = "") -> bool:
+        for line in code_content.split(os.linesep):
+            if not self.is_comment_line(line):
+                if 'import' in line and callee_package in line:
+                    if identifier:
+                        if identifier == callee_package:
+                            if all(word in line for word in ['import', callee_package]):
+                                return True
+                        else:
+                            if all(word in line for word in ['import', callee_package, 'as', identifier]) \
+                                    or all(word in line for word in ['from', callee_package, 'import', identifier]):
+                                return True
+                    else:
+                        if all(word in line for word in ['from', callee_package, 'import']):
+                            return True
+        return False
+


@tmihalac This one that you passing the code str instead of the document, and refactoring of callers were a good decision, it was simplified the readability of the code

zvigrinberg · 2025-11-03T10:44:44Z

src/vuln_analysis/utils/java_segmenters_with_methods.py

+        i += 1
+    return n - 1
+
+def _match_balanced_braces(src: str, i: int) -> int:


@tmihalac please document function purpose succinctly , what it gets as input, what it does, and what is the integer that it returns to callers

zvigrinberg · 2025-11-03T10:46:09Z

src/vuln_analysis/utils/java_segmenters_with_methods.py

+        i -= 1
+    return 0
+
+def _match_paren_backwards(src: str, i: int) -> int:


@tmihalac Function intention and purpose is not quite clear:
please document function purpose succinctly , what it gets as input, what it does, and what is the integer that it returns to callers

zvigrinberg · 2025-11-03T10:49:10Z

src/vuln_analysis/utils/java_segmenters_with_methods.py

+            i2 += 1
+        return ""
+
+    def _enum_constants_terminator(reg: _TypeRegion) -> int:


@tmihalac Function intention and purpose is not quite clear:
please document function purpose succinctly , what it gets as input, what it does, and what is the integer that it returns to callers

zvigrinberg · 2025-11-03T11:08:58Z

src/vuln_analysis/utils/java_utils.py

+
+    return s
+
+def convert_to_maven_artifact(package_name: str) -> str:


@tmihalac This is only used internally here, please refactor to start with underscore for a better readability and to reflect intention.
If you think that you're going to use it outside of this module in the future, you can leave it as is...

zvigrinberg · 2025-11-03T11:14:42Z

src/vuln_analysis/tools/transitive_code_search.py

    </tool_identity>
    <requires>
-        Input format: 'package_name,function_name' or 'package_name,class_name.method_name'
+        Input format: 'package_name,function_name' or 'package_name,class_name.method_name' or maven_gav(groupId:artifactId:version),fully_qualified_class_name.method_name


@tmihalac Did you try out the new package and function locator tool integration with JAVA?
is it working as expected?
This is something that needed to be checked manually, as we still haven't implemented integration tests yet.

How do I test it, isn't it called all the time ?

Looks to be working correctly, but haven't tried many inputs because I don't have

Hi OK.
All i'm asking is if it works as before or as good as it was before, with this new tool description.

For java it looks to be working the same, for go, c, python hard for me to test

Signed-off-by: Theodor Mihalache <[email protected]>

zvigrinberg · 2025-11-09T13:53:25Z

/retest vulnerability-analysis-on-pr

tmihalac · 2025-11-09T15:00:24Z

/retest vulnerability-analysis-on-pr

Signed-off-by: Theodor Mihalache <[email protected]>

…a-transitive-search # Conflicts: # src/vuln_analysis/tools/transitive_code_search.py

tmihalac · 2025-11-19T12:44:47Z

/retest vulnerability-analysis-on-pr

zvigrinberg

@tmihalac Good Job.
Please Take a look on my comments , most of them are minor ones.

Thanks.

src/vuln_analysis/utils/functions_parsers/java_functions_parsers.py

src/vuln_analysis/utils/java_chain_of_calls_retriever.py

src/vuln_analysis/utils/functions_parsers/java_functions_parsers.py

src/vuln_analysis/utils/java_utils.py

zvigrinberg · 2025-11-24T09:37:30Z

src/vuln_analysis/utils/java_utils.py

+        while True:
+            while i < n and s[i].isspace():
+                i += 1
+            if i < n and s[i] == '@':


@tmihalac Please replace @ with a constant JAVA_ANNOTATION_SYMBOL

zvigrinberg · 2025-11-24T09:58:24Z

src/vuln_analysis/utils/java_utils.py

+                out.append(x)
+        return out
+
+    def tup_dedup_keep_order(xs: List[Tuple[str, str]]) -> List[Tuple[str, str]]:


@tmihalac Is it possible to use a more meaningful argument name ? ( what is xs?)
In addition, would be better to document this function.

zvigrinberg · 2025-11-24T09:59:13Z

src/vuln_analysis/utils/java_utils.py

+    def simple_name_from_fqcn(fq: str) -> str:
+        return fq.split('.')[-1] if fq else fq
+
+    def dedup_keep_order(xs: List[str]) -> List[str]:


@tmihalac Is it possible to use a more meaningful argument name ? ( what is xs?)
In addition, would be better to document this function.

src/vuln_analysis/utils/java_utils.py

Signed-off-by: Theodor Mihalache <[email protected]>

…added docstring Signed-off-by: Theodor Mihalache <[email protected]>

…d docstring Signed-off-by: Theodor Mihalache <[email protected]>

Signed-off-by: Theodor Mihalache <[email protected]>

…st instead of set Signed-off-by: Theodor Mihalache <[email protected]>

Signed-off-by: Theodor Mihalache <[email protected]>

tmihalac requested a review from zvigrinberg October 5, 2025 14:46

tmihalac marked this pull request as ready for review October 5, 2025 14:46

zvigrinberg requested changes Oct 15, 2025

View reviewed changes

zvigrinberg reviewed Oct 16, 2025

View reviewed changes

tmihalac force-pushed the java-transitive-search branch from e9ee7f8 to 15e1ccf Compare October 29, 2025 17:50

tmihalac requested a review from zvigrinberg October 29, 2025 17:54

tmihalac force-pushed the java-transitive-search branch from f5fa123 to bc89d72 Compare October 30, 2025 10:37

tmihalac force-pushed the java-transitive-search branch 4 times, most recently from 8500b8d to 0197565 Compare November 2, 2025 13:10

zvigrinberg requested changes Nov 3, 2025

View reviewed changes

tmihalac force-pushed the java-transitive-search branch from 4ede2b8 to 4f62893 Compare November 3, 2025 12:49

Java transitive search

5ba4d90

Signed-off-by: Theodor Mihalache <[email protected]>

tmihalac force-pushed the java-transitive-search branch from 4f62893 to 5ba4d90 Compare November 9, 2025 13:04

tmihalac added 2 commits November 9, 2025 15:05

Java transitive search

a4ad0d5

Signed-off-by: Theodor Mihalache <[email protected]>

Java transitive search

6e8f9b3

Signed-off-by: Theodor Mihalache <[email protected]>

tmihalac requested a review from zvigrinberg November 18, 2025 14:36

tmihalac force-pushed the java-transitive-search branch from 134dca9 to 6e8f9b3 Compare November 19, 2025 09:27

tmihalac and others added 4 commits November 19, 2025 11:33

Java transitive search

8498113

Signed-off-by: Theodor Mihalache <[email protected]>

Java transitive search

8fb7924

Signed-off-by: Theodor Mihalache <[email protected]>

Java transitive search

eef3435

Signed-off-by: Theodor Mihalache <[email protected]>

Merge remote-tracking branch 'origin/java-transitive-search' into jav…

1489520

…a-transitive-search # Conflicts: # src/vuln_analysis/tools/transitive_code_search.py

zvigrinberg requested changes Nov 24, 2025

View reviewed changes

tmihalac added 6 commits November 24, 2025 12:47

search_for_called_function - moved inner methods to start

cb525be

Signed-off-by: Theodor Mihalache <[email protected]>

__determine_doc_package_name - added docstring

ea93c16

Signed-off-by: Theodor Mihalache <[email protected]>

introduced JAVA_ANNOTATION_SYMBOL constant

78de59e

Signed-off-by: Theodor Mihalache <[email protected]>

tup_dedup_keep_order changed parameter names to meaningful names and …

e76d7be

…added docstring Signed-off-by: Theodor Mihalache <[email protected]>

dedup_keep_order changed parameter names to meaningful names and adde…

96d5070

…d docstring Signed-off-by: Theodor Mihalache <[email protected]>

make_resolver - added docstring with explanation

5a51255

Signed-off-by: Theodor Mihalache <[email protected]>

tmihalac requested a review from zvigrinberg November 24, 2025 17:32

tmihalac added 2 commits November 26, 2025 20:23

Fix is_same_package for all parsers and fixed dict tree parents to li…

b2dbbbc

…st instead of set Signed-off-by: Theodor Mihalache <[email protected]>

Fix get_relevant_documents adding parents to list incorrectly

089d9e6

Signed-off-by: Theodor Mihalache <[email protected]>

	def set_input_for_next_run(git_repository: str, git_ref: str, included_extensions: list[str],
	excluded_extensions: list[str]):
	source_code_info = [SourceDocumentsInfo(type='code', git_repo=git_repository,
	ref=git_ref, include=included_extensions,
	exclude=excluded_extensions)]
	sbom_info_input = ManualSBOMInfoInput(packages=[SBOMPackage(name="a", version="1.0", system="blabla")])
	morpheus_input = AgentMorpheusInput(image=ImageInfoInput(source_info=source_code_info,
	sbom_info=sbom_info_input,
	analysis_type=AnalysisType.IMAGE),
	scan=ScanInfoInput(vulns=[VulnInfo(vuln_id="CVE-2025-1234")]))
	engine_input = AgentMorpheusEngineInput(input=morpheus_input, info=AgentMorpheusInfo())
	state: AgentMorpheusEngineState = AgentMorpheusEngineState(original_input=engine_input,
	code_vdb_path="",
	doc_vdb_path="",
	code_index_path="",
	cve_intel=[])
	ctx_state.set(state)


	async def get_transitive_code_runner_function():
	transitive_code_search = transitive_search(config=TransitiveCodeSearchToolConfig(), builder=None)
	async for function in transitive_code_search.gen:
	return function.single_fn

	async def test_transitive_search_golang_4():
	transitive_code_search_runner_coroutine = await get_transitive_code_runner_function()
	set_input_for_next_run(git_repository="https://github.com/openshift/oc-mirror",
	git_ref="b137a53a5360a41a70432ea2bfc98a6cee6f7a4a",
	included_extensions=["*/.go"],
	excluded_extensions=["go.mod", "go.sum", "*/test*.go"])
	result = await transitive_code_search_runner_coroutine("github.com/mholt/archiver,Unarchive")
	(path_found, list_path) = result

	assert path_found is True
	assert len(list_path) == 2

	def extract_method_name_with_params(self, java_src: str) -> str \| None:
	def extract_method_name_with_params(self, src: str) -> str \| None:

		def is_doc_type(self, doc: Document) -> bool:
		return doc.page_content.startswith("type")


		return s

		def convert_to_maven_artifact(package_name: str) -> str:

Java transitive search #130

Are you sure you want to change the base?

Java transitive search #130

Uh oh!

Conversation

tmihalac commented Oct 5, 2025

Uh oh!

zvigrinberg left a comment

Choose a reason for hiding this comment

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!

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!

tmihalac commented Oct 30, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

tmihalac commented Oct 30, 2025

Uh oh!

zvigrinberg commented Oct 30, 2025

Uh oh!

zvigrinberg 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!

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

tmihalac commented Oct 30, 2025 •

edited

Loading