Start using `g_autoptr` and `g_autofree` in all GLib based code ? (libuca, etc)

These are some Glib macros that tell the compiler to automatically free any Heap variable before a return statement.
With this, you no longer need to call g_free() or g_object_unref().

Only caveat : You need to use GCC or CLANG (and also all your variables have to initialized before hand, but this is only natural). Maybe a deal breaker ?

Really not urgent, could simply make the codebase more modern, memory safe and readable.

Use libpcap for 10Gb data transfer ?

Libpcap offers zero-copy data transfering for raw sockets (Shared memory between kernel space and user space). This is what is currently done in the driver, but it's messy and prone to errors for in the future.
Using libpcap would clean up the code and make it nore versatile while also maintaining performances (+ it's cross platform! ).
Only downside is that it adds a dependency to download...

Handle phantom specific features

I've added the full access to all of the phantom's features, e.g:

auto triggers based on image events
frame decimation
image cropping
video output settings
image calibration
etc

But they'd need some extra support on the code side of things. For example, the communication module doesn't take cropping or frame decimation into account. Something to think about in the future, based on specific needs.

Nested error management

Error management is already solid. But it would be nice to have a nested printing mechanism where each layer of the propagated errors was preceded by a newline+tab when printed.

Add levels of verbosity in phantom logs/warnings/messages

Glib mainly supports mainly the following: log, message, info, warning, error, debug.
It would be good to able to specify watch messages you wish to display accorrding to what you need.

Declutter uca-phantom by seperating phantom networking functions from actual camera class implementation

Proposal

Implement a seperate class/namespace that manages the phantom purely from a network POV. Use it to seperate the pure network functions from the overrided class functions of uca-phantom.c.

Automatic documentation generation

Readily available well-documented open-source documentation tools

Doxygen
Sphinx

Doxygen seems to more be widespread and have a better documetation. This could help us easily exhibit our public functions that should be used by the user and how they relate to the super class (which functions are overrided, which member variables are public, etc).
May also enable automatic UML style class-inheritance diagrams.

Work that would need to be done

Classes and their methods would need to have a commented header documenting the said-functions. A standard syntax would have to followed. Fortunately, this is already mostly done.

uca_phantom_get_variable freezes after too many immediate consecutive calls

Problem description

Behaviour is undefined when doing the following:

    GError *error = NULL;
    GValue val = G_VALUE_INIT;
    gboolean result = FALSE;

    UcaPhantomCommunicate *communicator = uca_phantom_communicate_new();
    gboolean connected = uca_phantom_communicate_attempt_connect(communicator, &error);

    if (!connected) {
        g_print ("Houston theres a problem: %s\n", error->message);
        g_error_free (error);
        g_object_unref (communicator);
        return FALSE;
    }

    struct timespec req, rem;
    req.tv_sec = 0;
    req.tv_nsec = 5e+6;

    for (int i=0; i < N_UNIT_PROPERTIES; i++) {
        result = uca_phantom_get_variable (communicator, i, &val, &error);
        if (!result && error != NULL) {
            g_print ("Houston we've got a problem: %s\n", error->message);
            g_error_free (error);
            g_object_unref (communicator);
            return FALSE;
        }
        print_gvalue (i, &val);
        g_value_unset (&val);
        // nanosleep(&req, &rem);
    }

    g_object_unref (communicator);

Indeed, after a certain and seemingly random amount of requests, the program feezes. However, by adding a blocking sleep command at the end of the loop, all requests are made and all answers are received.

Valgrind output

[...]
==45048== Process terminating with default action of signal 2 (SIGINT)
==45048==    at 0x4D01D7F: poll (poll.c:29)
==45048==    by 0x4AB12AD: g_socket_condition_timed_wait (in /usr/lib/x86_64-linux-gnu/libgio-2.0.so.0.7200.4)
==45048==    by 0x4AB161E: ??? (in /usr/lib/x86_64-linux-gnu/libgio-2.0.so.0.7200.4)
==45048==    by 0x4A8F1F6: g_input_stream_read (in /usr/lib/x86_64-linux-gnu/libgio-2.0.so.0.7200.4)
==45048==    by 0x10CDF9: uca_phantom_communicate (uca-phantom-communicate.c:517)
==45048==    by 0x10CDF9: uca_phantom_get_variable (uca-phantom-communicate.c:670)
==45048==    by 0x10B9B4: main (test_network.c:51)
==45048== 
==45048== HEAP SUMMARY:
==45048==     in use at exit: 278,549 bytes in 2,329 blocks
==45048==   total heap usage: 11,650 allocs, 9,321 frees, 9,845,914 bytes allocated
==45048== 
==45048== LEAK SUMMARY:
==45048==    definitely lost: 14 bytes in 1 blocks
==45048==    indirectly lost: 0 bytes in 0 blocks
==45048==      possibly lost: 3,216 bytes in 4 blocks
==45048==    still reachable: 250,999 bytes in 2,089 blocks
==45048==         suppressed: 0 bytes in 0 blocks
==45048== Rerun with --leak-check=full to see details of leaked memory
==45048== 
==45048== For lists of detected and suppressed errors, rerun with: -s
==45048== ERROR SUMMARY: 0 errors from 0 contexts (suppressed: 0 from 0)

The memory leak is likely due to the communicator object not being freed after the ctl+c signal.

Envision implementing unit tests for the phantom camera

Test network com with phantom

TODO

Test `Unit Structure` members as per the PH16 spec sheet

See spec sheet for description of each member variable. More variables are listed out than needed. Before implementing these unit tests, one must first verify which variables need to be known for the general functionning of uca-phantom, which need to have a default value set and which are prone to be modified.

info structure

Sensor information:

info.sensor
info.snsversion
info.cfa
info.filter

Versions and identification

Capabilities

Color correction (non concerned)

Nothing important to check

Camera status monitoring

info.modes

Meta stucture

meta.name, meta.lens, meta.fstop, meta.flen, meta.comment, meta.xset

Cam structure

Contains global user-alterable settings that aren’t tied to a particular cine.

Camera Sync Options

Global camera options

Eth structure

This section is probably quite important as it is directly related to the network settings of the camera.

Network settings

Video structure

General video output controls (are we concerned by this ? TODO: test these params out)

On screen display control

definitely not concerned by this

Video scaling

video.uzoom
video.vox, video.voy
video.vow, video.voh,
video.vw, video.vh

Video image adjustments (TODO: check if is relevant)

Irig structure

Structure containing the status of the camera timebase and range data acquisition unit. updated ~1Hz.

Mag (Cinemag) structure (important ?)

defc (default cine) structure

Variables that control the camera image capture process.

Cf structure (storage device) (important ?)

See spec sheet for flag vals

Cine tables structure (important ?)

Array of cine structures -> used to store images and meta data (e.g timestamps). More todo in understanding how these structures are layed out in memory.

Status variables

memory allocation

c#.start, c#.len
c#.frsize, c#.frspace

trigger time

c#.trigtime.secs, c#.trigtime.frac

acquisition parameters

See defc

cam/info/adj/meta substructures

These copies are done at the time the cine recording ends, and reflects the meta values.
in place at the time of recording.

Usets (important ?)

Probably not important

hw (Do not touch)

Auto structure ( important ?)

Add timestamps

Reading the timestamps is done the same way like the frames.
Possible implementations are:

Reading the timestamps of a cine before the first grab and then encoding them in the frames
Reading the timestamps at any point an make them accessible via a function or porperty

Frame rate not settable

Currently the frame rate is not settable. frames-per-second parameter reports 1/exposure-time (defined in uca base camera).

Modernizing uca-phantom to latest GObject standards and practices

Main objectives

Not much is actually needed to be done. The task would mostly entail

using GObject's net macro's for better typechecking, signal handling and error propagation.
Use GObject's new type registration methods. Most notably the private structure ones
These are currently not used.

important: check if updating uca-phantom to modern GObject will break compatibility with uca-camera. => I am reasonable certain that this is not an issue, but this should be sandbox-tested first.

ufo-kit / uca-phantom Goto Github PK

uca-phantom's Introduction

uca-phantom

Documentation

User information

Changing the acquisition mode

Transfering data with a 10GE connection

Testing with the mock software emulation

Developer information

uca-phantom's People

Contributors

Watchers

uca-phantom's Issues