ericsson / codechecker Goto Github PK

It should be possible to quickly check a small project (set of files) for bugs without storing the results into the postgres db.

The defect list should appear on the console without the bug paths, only showing the source location of the last step of the bugpath, checker name, bug message.

Only the build command should be the required command line parameter:

CodeChecker quickcheck -b "make"

Do not enable suppress on the web GUI if no suppress file was given to the server.

Add delete run results support to the web gui.

The failed check command (build action) should be also printed to the dump file header.

Change plist parser to handle the plist output generated by the bug hash patch.
(mainly bug hash and bug hash type tags)
http://reviews.llvm.org/D10305

CodeChecker viewer doesn't work in Internet Explorer 9 (I didn't try with another IE).

when running CodeChecker check from multiple hosts in a distributed way, you don't know which command will be executed first.

create documentation for the importnat functions/structures

CodeChecker checkers

command should print the list of checkers which are on by default

If we close and re-open a run or a diff, we are unable to open a FileView to any of the reports.

The addChild function fails on the OverviewTC for some reason.

Reproducable, the error in the browser console is as follows:
"Uncaught TypeError: Cannot read property 'appendChild' of null"

For example, the program that I am trying to build has this define:
-DDEFAULT_ICU_PLUGINS="/usr/local/lib/icu"\

Codechecker will fail this:
[ERROR] - The compile database is not valid.
[ERROR] - Invalid \escape: line 29848 column 786 (char 10574627)

Thanks for this tool! Looking awesome right now 👍

My aim is to provide a skip file for CodeChecker within my project's source tree. But I want to only parse my own code and not optional and potential 3rd-party libraries. Thus, I can not put in absolute paths such as /usr/include. As well it's not possible to exclude specific files in the project's source tree (e.g. <PROJECT>/include/3rdparty) as I can not make any assumptions on the location of the project tree.

I tried specifying patterns such as eigen3/, */eigen3, /boost or */boost in a skip file. But that did not work. Only absolute paths such as /usr/include/eigen3 are working.

It would be awesome being able to exclude paths based on patterns or - even better - whitelist only specific files or paths only in the project's source tree.

The following annotation format should be applied above the line where the bug was reported.
The list should contain checker names or the 'all' keyword if all bugs should be suppressed during the analysis.

// codechecker_suppress[checker_name1, checker_name2] reason for suppress
// is that checker is faulty.
int i=0;
or
// codechecker_suppress[all] reason for suppress

The "reason for suppress" field should be given by the user and it is a free text that can contain any character. Only whitespace characters are not accepted. If the "reason for suppress" field is missing, a warning message will be generated in the suppress_comment the bug database.
Reasons specified in multiple lines are accepted.

Show in the run overview page, in the checker filter dropdown in parenthesis how many hits are in the run, per checker.

Also show the same information for bug priority and suppressed, unsuppressed list.

check https://github.com/Ericsson/codechecker/blob/master/docs/user_guide.md

Write a user guide for quickcheck command.

show how many reports are shown by a filtered request

sometimes it is useful to specify in a file the directories
from where files need to be checked.

this is especially problematic if the project takes file from many places of the filesystem.

CodeChecker check -f config.file is supported. Please describe the format.

if the build command is logged with the same parameters multiple times, do not recheck the file

A "bug" should be skipped if its path's last point/step should be skipped according to a skip file (given by --skip command line option).

and not id field

Introduce a command line parameter
--clang /path/to/clang/binary

to specify to codechecker which clang binary should be used.
If this parameter is not specified, then the path specified config.json file should be used.

and in tooltip show the also the error summary.

today a bug is listed as many times as it is reported by a build action.
a bug in a header file, which is compiled many times will be reported many times.

We should check if the browser the gui is opened in is compatible with the WEBGUI and its dependencies (dojo, codemirror, jsplumb, jquery, etc.)

show in a popup the minimum required browser version only in case the used browser is not compatible with codechecker.

it would be useful if the important views of the web GUI would have permalink URLs.
the following should have permalinks:-
-run-list
-bug list overview including filters
-view of a concrete defect
-a concrete page of the documentation and user guide

Then it would be possible the to send the the defect reports in email or use the browser back button.

Set version to 5.0

There are several cases when the command prints an error but the command returns with status code 0. It should always return with a non-zero status code in case of error to make automation of the commands easier.

Currently the sort function of the GUI only sorts the defects on the current page, however it should perform an ordering on the whole bug list in the project.

It should be possible to specify a config file to
CodeChecker check subcommand
where the checker specific configuration can be added.

Preferably in json format:

[  
   {  
      "checker_name":[  
         {  
            "parameter_name":"value"
         },
         {  
            "parameter_name":"value"
         }
      ]
   }
]

For example:

[  
   {  
      "NonPrivateDataChecker":[  
         {  
            "exclude_structs":"true"
         }
      ]
   }
]

introduce an env variable to the logger to log only the execution of elf binaries (gcc, g++ are binaries)

the name of the env variable should be 🎱
CC_LOG_BIN_ONLY
default 1

Codechecker check sumbcommand parameters are not properly described and skip file and suppress file doc do not belong to debug mode in the user_guide.md

debug mode section should be erased as it has nothing to do with debugging.

Codechecker check crashed with the following:

Traceback (most recent call last):
File "/Codechecker/cc_bin/CodeChecker", line 28, in
with open(original_env_file, 'wb') as env_save:
IOError: [Errno 13] Permission denied: '/Codechecker/cc_bin/../config/original_env.pickle'

Proposed Solution:
The binary package directory in general is not writable.
Use the /tmp directory for temp files.
for example /tmp/codechecker_[username]

CodeChecker does not report findings from all source files if a compiler command contains multiple files.

Example:

g++ source1.cpp source2.cpp source3.cpp source4.cpp

it would be great to receive an email when new issues are found by the tool

CodeChecker can only run Clang SA, but there are many checkers implemented in Clang Tidy.

Codechecker shall be able to wrap
-config parameters
-checker enable/disable list
and parse the output of Clang tidy in exactly the same format as Clang SA

when showing run overview the checker names that are active in the current run are loaded synchronously. this couses the web browser to hang when there are many thousand hits.

introduce a new configuration parameter to
CodeChecker check
subcommand to pass through parameters to clang.

-e Pass extra parameters to clang.

this could be useful to specify parameters such as -analyzer-opt-analyze-headers or -analyzer-max-loop etc.

Codechecker check time increases non-linearly with the increasing run number/database size

when we add a new hit in the database, we check the uniqueness of the report based on the hash alone.
if there is a hash clash, then we do not add the report to the database.
however this could lead to missing reports when there is a hash clash. it is a problem especially when the bug is reported in the global namespace, because there is no named declaration context (and the filenanme is not in the hash).

check uniqueness based on hash first, then if there is a clash, check the checker name, file name and bug position.

currently codechecker by default does not clean up the temporary plist files after the check, only if
'-c' clean argument is set
this should be changed to a new default behaviour:
by default the temporary files should be deleted during the check and only kept if it is explicitly set in the command line by a keep-temporary files option

it would be useful to be possible to give command line parameters as env variables
this would make the command line shorter.

if command line is specified it should override env var and default settings:
default->env-var->command line (override sequence)

in case of enable/disable checker list: the default should be extended by either the env var or the command line param

these params should have env var equivalent:
-workspace
-dbport
-dbhost
-dbuser
-dbname
-enable/disable checker list
-viewport
-suppress file path

if the env var is speficied then the CodeChecker script should list their values when started.

When the report database is large, getting report details takes a long time and the report server uses gigabytes of memory.

now can only be seen in debug mode.

[2048] <139634032555776> - report_server.run_server() [CC SERVER] - Creating new database schema done in 0
[2048] <139634032555776> - report_server.run_server() [CC SERVER] - Version mismatch. Expected database version: v4.0
[2048] <139634032555776> - report_server.run_server() [CC SERVER] - Version from the database is: v3.0
[2048] <139634032555776> - report_server.run_server() [CC SERVER] - Please update your database.
[1997] <139634032555776> - client.start_server() [CLIENT] - Waiting for checker server to start.
[1997] <139634032555776> - client.start_server() [CLIENT] - Waiting for checker server to start.
[1997] <139634032555776> - client.start_server() [CLIENT] - Waiting for checker server to start.
[1997] <139634032555776> - client.start_server() [CLIENT] - Failed to start checker server.

today it is ordered by filename only, but within the same file it is random.

database migration feature is implemented, however it is undocumented.
provide developer documentation on how to extend the codechecker db schema so that it supports upgrades.

it should be possible to remove any filter from the run overview filter list