Limit class loading #604

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Merged

kroening merged 6 commits into diffblue:master from mgudemann:limit_class_loading

Mar 13, 2017

Contributor

mgudemann commented Mar 4, 2017 •

edited

Loading

This PR adds the capability to load only the subset of class files from a jar which match a regex specified via the java-cp-include-files parameter.
requires #548

mgudemann force-pushed the limit_class_loading branch from 64b5643 to 346d57e Compare

March 4, 2017 13:18

Member

peterschrammel commented Mar 4, 2017

What's the advantage of that in comparison to lazy class loading (#407)?

Contributor Author

mgudemann commented Mar 4, 2017

It's in addition to lazy loading. If I understand correctly, lazy loading will only load the parts that are actually used and not everything that is present. This allows you to limit beforehand what is seen from a jar file, @marek-trtik had asked for such a feature.

Member

peterschrammel commented Mar 4, 2017

I see. These two are orthogonal.

mgudemann mentioned this pull request

WAR load support #608

Closed

Member

peterschrammel commented Mar 7, 2017

cegis needs json.a to compile: https://travis-ci.org/diffblue/cbmc/jobs/208614294#L2370

mgudemann force-pushed the limit_class_loading branch from fb4917f to 8e96c2a Compare

March 8, 2017 08:58

Contributor Author

mgudemann commented Mar 8, 2017

added json.a to cegis Makefile

tautschnig requested changes

View reviewed changes

Collaborator

tautschnig left a comment

Various requests for cleanup and logic simplifications.

src/java_bytecode/jar_file.cpp Outdated

    
              #include <sstream>

              #include <json/json_parser.h>

              #include <unordered_set>

Collaborator

tautschnig Mar 9, 2017

Nit picking: group STL headers, only then start with local headers

Contributor Author

mgudemann Mar 10, 2017

done

src/java_bytecode/jar_file.cpp Outdated

    
              #include "jar_file.h"

              #include <util/suffix.h>

              #include <iostream>

Collaborator

tautschnig Mar 9, 2017

Is iostream really required?

Contributor Author

mgudemann Mar 10, 2017

done

src/java_bytecode/jar_file.cpp Outdated

    
                         get_message_handler(),

                         json_cp_config))

                      throw "cannot read JSON input configuration for JAR loading";

                    assert(json_cp_config.is_object() && "JSON has wrong format");

Collaborator

tautschnig Mar 9, 2017

This is not an internal error (it's invalid user input), I think - should be throw instead of assert.

Contributor Author

mgudemann Mar 10, 2017

done

src/java_bytecode/jar_file.cpp Outdated

    
                      throw "cannot read JSON input configuration for JAR loading";

                    assert(json_cp_config.is_object() && "JSON has wrong format");

                    jsont include_files=json_cp_config["classFiles"];

                    assert(include_files.is_array() && "JSON has wrong format");

Collaborator

tautschnig Mar 9, 2017

As above: invalid user input, not an internal error.

Contributor Author

mgudemann Mar 10, 2017

done

src/java_bytecode/jar_file.cpp Outdated

    
                      add_file|=std::regex_match(file_name, string_matcher, regex_matcher);

                    // load .class file only if it is in the match set

                    else

                      add_file|=set_matcher.count(file_name)>0;

Collaborator

tautschnig Mar 9, 2017

Why use count() instead of find?

Contributor Author

mgudemann Mar 10, 2017

because it was shorter, also it is an unordered map so it shouldn't be less efficient as count can only be 0 or 1 and if an element is found, the implemention likely stops traversing the unordered map. In any case, I change this to find() to be consistent with the code base.

src/java_bytecode/jar_file.h Outdated

    
                jar_filet &operator()(const std::string &file_name)

                {

                  assert(!java_cp_include_files.empty() && "class regexp cannot be empty");

Collaborator

tautschnig Mar 9, 2017

This is a user input, I believe?! Should be thrown, not asserted.

Contributor Author

mgudemann Mar 10, 2017

done

src/java_bytecode/jar_file.h

    
                  file_mapt::iterator it=file_map.find(file_name);

                  if(it==file_map.end())

                  {

                    jar_filet &jar_file=file_map[file_name];

                    jar_file.open(file_name);

                    jar_file.set_message_handler(get_message_handler());

Collaborator

tautschnig Mar 9, 2017

Does jar_filet benefit from being a messaget (see comments above)?

Contributor Author

mgudemann Mar 10, 2017

I'll add a status output to see about hidden files, also the JSON parser needs a valid one

src/java_bytecode/java_bytecode_language.cpp

    
            @@ -55,6 +56,29 @@ void java_bytecode_languaget::get_language_options(const cmdlinet &cmd)
          
                  lazy_methods_mode=LAZY_METHODS_MODE_CONTEXT_INSENSITIVE;

                else

                  lazy_methods_mode=LAZY_METHODS_MODE_EAGER;

Collaborator

tautschnig Mar 9, 2017

Please add a blank line. Codewithoutblanksisreallyhardtoread.

src/java_bytecode/java_bytecode_language.cpp Outdated

    
                       get_message_handler(),

                       json_cp_config))

                    throw "cannot read JSON input configuration for JAR loading";

                  assert(json_cp_config.is_object() && "JSON has wrong format");

Collaborator

tautschnig Mar 9, 2017

See above for why I believe these aren't internal errors.

src/java_bytecode/java_class_loader.h Outdated

    
            @@ -20,6 +20,11 @@ class java_class_loadert:public messaget
          
              {

              public:

                java_bytecode_parse_treet &operator()(const irep_idt &);

                void set_java_cp_include_files(std::string &java_cp_include_files)

Collaborator

tautschnig Mar 9, 2017

A blank line, please.

Matthias Güdemann added 4 commits

March 9, 2017 18:30


          allow regex to limit class files loaded from JAR

b236ee4


          support class load configuration via JSON file

ed34a73

e.g.

```
{
  "jar":
  [
    "A.jar",
    "B.jar"
  ],
  "classFiles":
  [
    "jarfile3$A.class",
    "jarfile3.class"
  ]
}
```


          add jars from JSON config to classpath

469897f


          add regression test

3d009a9

mgudemann force-pushed the limit_class_loading branch from 8e96c2a to 1975394 Compare

March 10, 2017 10:19

Matthias Güdemann added 2 commits

March 10, 2017 14:31


          take comments into account

b473638


          change from vector indices to map

53fad70

mgudemann force-pushed the limit_class_loading branch from eae1b8b to 53fad70 Compare

March 10, 2017 13:31

tautschnig approved these changes

View reviewed changes

peterschrammel approved these changes

View reviewed changes

peterschrammel assigned kroening

peterschrammel added enhancement Java Front End labels

kroening merged commit 36c2e2d into diffblue:master

NathanJPhillips pushed a commit to NathanJPhillips/cbmc that referenced this pull request


          Merge pull request diffblue#604 from diffblue/jd/feature/pipeline_sto…

b3e1d03

…p_on_exception

[SEC-677] Turning DI off by default and making errors fail pipeline

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

enhancement Java Front End