gyrovorbis / libgimbal Goto Github PK

C17-based extended standard library, cross-language type system, and unit testing framework targeting Sega Dreamcast, Sony PSP and PSVita, Windows, Mac, Linux, Android, iOS, and WebAssembly.

Home Page: http://libgimbal.elysianshadows.com

License: MIT License

CMake 1.03% C 92.27% C++ 6.69%

algorithms allocators c containers core dreamcast object-oriented-c type-system unit-testing utility-library

libgimbal's People

Contributors

Stargazers

Watchers

Forkers

aaronglazer thesubmarinedoorgunner-forks suicvne cualquiercosa327

libgimbal's Issues

Add XDG Support

Add support for reading/writing data, config files, etc. using the XDG spec: https://specifications.freedesktop.org/basedir-spec/basedir-spec-latest.html

https://stackoverflow.com/questions/43853548/xdg-basedir-directories-for-windows/49783629#49783629

Add signals for catching failed GblTestScenarios and GblTestSuites

Thanks to @suicvne for raising this issue:

It is often desirable for the user to be able to capture or filter over all of the failing test cases when executing a series of GblTestSuites from a GblTestScenario. You can currently do this by:

Overriding GblTestSuiteClass::pFnCaseRun() to capture the result code.
Iterating over the test suites within a test scenario and logging the failing issue:

for(int i = 0; i < GblObject_childCount(GBL_OBJECT(pScenario)); ++i) {
    GblTestSuite* pSuite = GblObject_findChildByIndex(GBL_OBJECT(pScenario), i);
    if(pSuite->casesFailed) {
        printf("%s has reested with code %s at %s:%s:%zu\n",
               GblObject_name(GBL_OBJECT(pSuite)), gblResultString(pSuite->failingIssue.result),
               pSuite->failingIssue.sourceLocation.pFile, pSuite->failingIssue.sourceLocation.pFunc,
               pSuite->failingIssue.sourceLocation.line);
    }
}

Unfortuantely there are a few shortcomings:

There is no way to retrieve the name of or function pointer to the exact test case that failed.
This would be SO MUCH MORE PRISTINE from an API perspective using the signal/slots mechanism to emit a signal when a test case fails.

Make a way to have this install itself after building or something

Like make install but whatever the cmake way is or whatever, idc.

UUIDs are not actually 128-bits of randomness

Problem

In gimbal_uuid.c, UUIDs are generated with the following:

// Implemented per RFC4122 section 4.4
GBL_EXPORT GBL_RESULT GblUuid_initV4(GblUuid* pSelf) {
    GBL_CTX_BEGIN(NULL);
    GBL_CTX_VERIFY_POINTER(pSelf);

    // First initialize all bytes to random/psuedo-random numbers
    for(unsigned b = 0; b < GBL_UUID_BYTE_COUNT; ++b) {
        pSelf->bytes[b] = gblRandRange(0, 255);
    }

where gblRandRange(0,255) resolves to

GBL_EXPORT int gblRandRange(int min, int max)  {
    GBL_ASSERT(max <= RAND_MAX);
    static GblBool seeded = GBL_FALSE;
    if(!seeded) GBL_UNLIKELY {
        srand((unsigned)gblSeed(0));
        seeded = GBL_TRUE;
    }
    return (rand() % (max - min + 1)) + min;
}

The C standard library rand() will typically only have a $2^{32} -1$ long period before it begins repeating for any given seed.
The issue arises, because generating multiple random numbers, and concatenating them, does not increase randomness. Since a PRNG's output is determinate, for any given seed, every next value is determinate.

Also, there is no reseeding after it is initially seeded.

In the following program I wrote which basically implements the same behavior

#include <stdio.h>
#include <stdlib.h>
#include <time.h>
#include <string.h>
#include <stdint.h>

#define SIZE 16

int main(void)
{
    srand(time(NULL));
    
    uint8_t reference[SIZE] = {0};
    uint8_t current[SIZE] = {0};
    
    printf("Reference: ");
    for(int i = 0; i < SIZE; i++)
    {
        reference[i] = (uint8_t)(rand() % 256);
        printf("%02X ", reference[i]);
    }
    printf("\nFinding duplicates...\n");
    
    clock_t start = clock();

    for(uint64_t i = 1; i < UINT64_MAX; i++)
    {
        for(int j = 0; j < SIZE; j++)
            current[j] = (uint8_t)(rand() % 256);
        
        
        if(!memcmp(reference, current, SIZE))
        {
            printf("Duplicate found after %llu generations\n", i);
            printf("Duplicate: ");
            for(int k = 0; k < SIZE; k++) printf("%02X ", current[k]);
            printf("\n");
            break;
        }
    }
    
    clock_t end = clock();
    float sec = (float)(end-start) / CLOCKS_PER_SEC;
    
    printf("CPU time: %.3f seconds.\n", sec);
    
    return 0;
}

Gives the following output

Reference: 51 82 D7 92 DF 69 AB B6 D1 B4 A1 8E 60 0E C8 B9
Finding duplicates...
Duplicate found after 1048576 generations
Duplicate: 51 82 D7 92 DF 69 AB B6 D1 B4 A1 8E 60 0E C8 B9
CPU time: 0.157 seconds.

Solution

To generate proper UUIDs with all the necessary bits of randomness, implementing a 128-bit PRNG yourself will likely be the best option. An LFSR or an LCG are commonly used in PRNGs.

In addition, you could also find out a scheme for reseeding your RNG in-between number generations, that could help too.

[Bug] Unaligned GblTestFixture/GblPrivateData

Found an issue with unaligned memory access on Dreamcast for the unit tests.

When creating a test fixture with:

GBL_TEST_FIXTURE {
    char dummy;
};

The Dreamcast will shit itself upon creation of the GblTestFixture instance spewing out event code 0x100 which Falco tracked down to represent an "unaligned memory write".

The output from dc-tool-ip:

KallistiOS Git revision 34f7e943-dirty:
  Fri May 31 23:59:48 EDT 2024
  [email protected]:/opt/toolchains/dc/kos
  sh-elf-gcc (GCC) 14.1.0
thd: pre-emption enabled, HZ=100
maple: active drivers:
    Dreameye (Camera): Camera
    Sound Input Peripheral: Microphone
    PuruPuru (Vibration) Pack: JumpPack
    VMU Driver: Clock, LCD, MemoryCard
    Mouse Driver: Mouse
    Keyboard Driver: Keyboard
    Controller Driver: Controller
    Lightgun: LightGun
  DMA Buffer at ac4509e0
vid_set_mode: 640x480 VGA with 1 framebuffers.
dc-load console support enabled
maple: attached devices:
  A0: Dreamcast Controller          (01000000: Controller)

* =========== Cascade Information ===========
*   Version:      0.3.0
*   Git Hash:     
* =========== Cascade Information ===========
* pScenario: 0x8c469144
* Enqueue Central Dispatch Test
* Enqueue Logging Test
Unhandled exception: PC 8c2296b0, code 1, evt 0100
 R0-R7: 00000001 00000000 8cfff8dc 0000003f 00000000 00000000 00000001 00000000
 R8-R15: 8cfff860 8c471398 8c471300 8c47141d 8cfff91c 00000000 8cfff954 8cfff854
 SR 40000000 PR 8c22980c
-------- Stack Trace (innermost first) ---------
   8cfff95c
   (invalid frame pointer)
-------------- End Stack Trace -----------------
kernel panic: unhandled IRQ/Exception
arch: aborting the system

Fix on my code was to do this:

GBL_TEST_FIXTURE {
    char dummy[4];
};

We tracked down where in the code the problem was happening and found code was able to proceed further once this was fixed.

gyrovorbis / libgimbal Goto Github PK

libgimbal's People

Contributors

Stargazers

Watchers

Forkers

libgimbal's Issues

Add XDG Support

Add signals for catching failed GblTestScenarios and GblTestSuites

Make a way to have this install itself after building or something

UUIDs are not actually 128-bits of randomness

Problem

Solution

[Bug] Unaligned GblTestFixture/GblPrivateData

Recommend Projects

React

Vue.js

Typescript

TensorFlow

Django

Laravel

D3

Recommend Topics

javascript

web

server

Machine learning

Visualization

Game

Recommend Org

Facebook

Microsoft

Google

Alibaba

D3

Tencent