Smalltalk › Squeak › Squeak - Dev

<CDECL kerfuffled on OpenGL. was (Re: OGLUnixX11LE >> glPixelStorei: with:

_‹ Previous Topic Next Topic _›

Classic

List

Threaded

7 messages Options

Squeak - Dev mailing list

<CDECL kerfuffled on OpenGL. was (Re: OGLUnixX11LE >> glPixelStorei: with:

Ok, progress....

SystemNavigation default browseAllSelect: [:m| m pragmas notEmpty] localToPackage: #OpenGL

SystemNavigation default browseAllSelect: [:m| m primitive isZero and: [m pragmas notEmpty]] localToPackage:#OpenGL

Its only showing primitive 120, it doesn't recognize the <apicall:module> as a pragma.

Here is an example of it not being recognized as a pragma vs one that is:

Pragma allNamed:#generated in:FFITestPoint2

Pragma allNamed:#apicall:module: in:OpenGL

Reading the class comment of OGLExtManager now trying to grok what's going wrong. It has some code to rewrite a method behind the scenes.

Whatever is rewriting the methods based on the <apicall:module:> pragma probably cannot see it.

Or, maybe I am completely in the wrong ballpark.

cool code, though. (:

Beckmann, Tom

Re: <CDECL kerfuffled on OpenGL. was (Re: OGLUnixX11LE >> glPixelStorei: with:

Could you describe the actual problem you're currently facing? If you successfully got a triangle to display, you're most likely good to go. If you were trying to use an OpenGL Extension method (typically recognizable by an ARB or EXT suffix) you will go through the OGLExtManager, which has to resolve function pointers to extension methods at runtime. This might be another source of problems, but for regular OpenGL 1 usage for immediate mode drawing (glVertex... and friends) you should be good to go once the first triangle has popped up :)

Best,
Tom
________________________________________
From: Squeak-dev <[hidden email]> on behalf of gettimothy via Squeak-dev <[hidden email]>
Sent: Sunday, May 10, 2020 9:33:21 PM
To: squeak-dev
Subject: [squeak-dev] <CDECL kerfuffled on OpenGL. was (Re: OGLUnixX11LE >> glPixelStorei: with:

Ok, progress....
SystemNavigation default browseAllSelect: [:m| m pragmas notEmpty] localToPackage: #OpenGL

SystemNavigation default browseAllSelect: [:m| m primitive isZero and: [m pragmas notEmpty]] localToPackage:#OpenGL

Its only showing primitive 120, it doesn't recognize the <apicall:module> as a pragma.

Here is an example of it not being recognized as a pragma vs one that is:

Pragma allNamed:#generated in:FFITestPoint2

Pragma allNamed:#apicall:module: in:OpenGL

Reading the class comment of OGLExtManager now trying to grok what's going wrong. It has some code to rewrite a method behind the scenes.

Whatever is rewriting the methods based on the <apicall:module:> pragma probably cannot see it.

Or, maybe I am completely in the wrong ballpark.

cool code, though. (:

Levente Uzonyi

Re: <CDECL kerfuffled on OpenGL. was (Re: OGLUnixX11LE >> glPixelStorei: with:

In reply to this post by Squeak - Dev mailing list

On Sun, 10 May 2020, gettimothy via Squeak-dev wrote:

> Ok, progress....
> SystemNavigation default browseAllSelect: [:m| m pragmas notEmpty] localToPackage: #OpenGL
>
> SystemNavigation default browseAllSelect: [:m| m primitive isZero and: [m pragmas notEmpty]] localToPackage:#OpenGL
>
>
>
> Its only showing primitive 120, it doesn't recognize the <apicall:module> as a pragma.

FFI calls are not pragmas. They support different syntax, so they are
language extensions.

Levente

>
> Here is an example of it not being recognized as a pragma vs one that is:
>
> Pragma allNamed:#generated in:FFITestPoint2
>
> Pragma allNamed:#apicall:module: in:OpenGL
>
>
>
> Reading the class comment of OGLExtManager now trying to grok what's going wrong. It has some code to rewrite a method behind the scenes.
>
>
> Whatever is rewriting the methods based on the <apicall:module:> pragma probably cannot see it.
>
> Or, maybe I am completely in the wrong ballpark.
>
> cool code, though. (:
>
>
>
>
>

Squeak - Dev mailing list

Re: <CDECL kerfuffled on OpenGL. was (Re: OGLUnixX11LE >> glPixelStorei: with:

---- On Sun, 10 May 2020 16:04:35 -0400 Levente Uzonyi <[hidden email]> wrote ----

FFI calls are not pragmas. They support different syntax, so they are
language extensions.

Levente

thx,

makes sense. in terms of FFI the <generated> is just documentation/marker stuff for use by the FFI test . Those classes, themselves are not Tests.

Here is my current thinking, there are x possiblities.

possibility 1. The <apicall:module> is a marker/tag used by "something" to insert platform specific method code . I see this approach in OGLExtManager loadExtension:

in this possibility, then the fact that it is not recognized as a Pragma would explain the failure to replace the code.

possiblity 2. the <apicall:module:> is an actual call

if this is the case, it does not work, and I need a butt-simple library call to test this possibility.

Or, I need to find all existing <apicall:module:> methods and see if they work.

thank you for your time

codefrau

Re: <CDECL kerfuffled on OpenGL. was (Re: OGLUnixX11LE >> glPixelStorei: with:

On Sun, May 10, 2020 at 3:21 PM gettimothy via Squeak-dev <[hidden email]> wrote:

---- On Sun, 10 May 2020 16:04:35 -0400 Levente Uzonyi <[hidden email]> wrote ----

FFI calls are not pragmas. They support different syntax, so they are
language extensions.

Levente

thx,

makes sense. in terms of FFI the <generated> is just documentation/marker stuff for use by the FFI test . Those classes, themselves are not Tests.

Here is my current thinking, there are x possiblities.

.

possibility 1. The <apicall:module> is a marker/tag used by "something" to insert platform specific method code . I see this approach in OGLExtManager loadExtension:
in this possibility, then the fact that it is not recognized as a Pragma would explain the failure to replace the code.

possiblity 2. the <apicall:module:> is an actual call
if this is the case, it does not work, and I need a butt-simple library call to test this possibility.
Or, I need to find all existing <apicall:module:> methods and see if they work.

thank you for your time

As we mentioned before, get the plain FFI test working before you try *anything* with OpenGL.

- Vanessa -

Squeak - Dev mailing list

Re: <CDECL kerfuffled on OpenGL. was (Re: OGLUnixX11LE >> glPixelStorei: with:

all FFI tests pass. In an earlier email I gave ab ecmple of incoking an FFI call that uses the cdecl

---- On Sun, 10 May 2020 19:33:26 -0400 [hidden email] wrote ----

On Sun, May 10, 2020 at 3:21 PM gettimothy via Squeak-dev <[hidden email]> wrote:
---- On Sun, 10 May 2020 16:04:35 -0400 Levente Uzonyi <[hidden email]> wrote ----

FFI calls are not pragmas. They support different syntax, so they are
language extensions.

Levente

thx,

makes sense. in terms of FFI the <generated> is just documentation/marker stuff for use by the FFI test . Those classes, themselves are not Tests.

Here is my current thinking, there are x possiblities.

.

possibility 1. The <apicall:module> is a marker/tag used by "something" to insert platform specific method code . I see this approach in OGLExtManager loadExtension:
in this possibility, then the fact that it is not recognized as a Pragma would explain the failure to replace the code.

possiblity 2. the <apicall:module:> is an actual call
if this is the case, it does not work, and I need a butt-simple library call to test this possibility.
Or, I need to find all existing <apicall:module:> methods and see if they work.

thank you for your time

As we mentioned before, get the plain FFI test working before you try *anything* with OpenGL.

- Vanessa -

codefrau

Re: <CDECL kerfuffled on OpenGL. was (Re: OGLUnixX11LE >> glPixelStorei: with:

Ah, great. And since in another mail you mentioned that even CroquetGL works there I'd recommend to just look into that running system whenever something seems odd - quite often it may turn out to be a red herring, like the apicall issue.

- Vanessa -

On Mon, May 11, 2020 at 4:19 AM gettimothy <[hidden email]> wrote:

all FFI tests pass. In an earlier email I gave ab ecmple of incoking an FFI call that uses the cdecl

---- On Sun, 10 May 2020 19:33:26 -0400 [hidden email] wrote ----

On Sun, May 10, 2020 at 3:21 PM gettimothy via Squeak-dev <[hidden email]> wrote:
---- On Sun, 10 May 2020 16:04:35 -0400 Levente Uzonyi <[hidden email]> wrote ----

FFI calls are not pragmas. They support different syntax, so they are
language extensions.

Levente

thx,

makes sense. in terms of FFI the <generated> is just documentation/marker stuff for use by the FFI test . Those classes, themselves are not Tests.

Here is my current thinking, there are x possiblities.

.

possibility 1. The <apicall:module> is a marker/tag used by "something" to insert platform specific method code . I see this approach in OGLExtManager loadExtension:
in this possibility, then the fact that it is not recognized as a Pragma would explain the failure to replace the code.

possiblity 2. the <apicall:module:> is an actual call
if this is the case, it does not work, and I need a butt-simple library call to test this possibility.
Or, I need to find all existing <apicall:module:> methods and see if they work.

thank you for your time

As we mentioned before, get the plain FFI test working before you try *anything* with OpenGL.

- Vanessa -