Use native build instead of x86/jetson target #31

lida2003 · 2024-11-25T23:29:36Z

Discussion: Ignore msposd_jetson for gs on jetson orin #29

lida2003 · 2024-11-25T23:38:09Z

sudo apt-get update
sudo apt-get upgrade

henkwiedig · 2024-11-26T06:47:31Z

I like this more !

Can you change .github/workflows/build.yml as well to match this ?

lida2003 · 2024-11-26T08:48:08Z

@henkwiedig Please review (I just do it first time, consulting https://docs.github.com/en/actions), I don't know how to trigger the test. But those script and binary name should be simple for review.

And PLEASE send me a link if there is any steps for local test.

henkwiedig · 2024-11-26T19:12:41Z

Testbuild looks fine, hi3536dv100_fpv doesn't build locally as well with main. So no regression for that.
When you want to locally run msposd on x86 you will need receiving hardware, setup the msp stream to receive from the drone. Just like on jetson. Then use gstreamer to start the video and overlay it with msposd. The screenshot you postet is what you will get when msposd is runnig and no mspdata is received. So this is expected with no forther config.

What about the DISPLAY envvar ? I now testest more then 3 ways to set this on remote ssh session.
ssh ... 'export DISPLAY=:0; msposd ...
ssh -o SendEnv=DISPLAY root@.... msposd....
edit .bashrc
The way it is now we completely disable other users to set the DISPLAY as they wish.
Unless there is really good reason for this i think this should be removed.

lida2003 · 2024-11-26T21:57:26Z

If you don't use it, then a segfault when launching from ssh, and no warning, no feedback.

henkwiedig · 2024-11-27T06:25:54Z

Yes, but why setting this in the application, especially forcefully overwrite it ?
You can use one of the million other way to set the DISLPLAY ?
There might be setups with two displays, one for Video one for telemetry/mission plan.
The user will not have the option to choose which is which.

lida2003 · 2024-11-27T06:38:46Z

Yes, but why setting this in the application, especially forcefully overwrite it ?

Not forcefully.

You can use one of the million other way to set the DISLPLAY ? There might be setups with two displays, one for Video one for telemetry/mission plan. The user will not have the option to choose which is which.

That's what I consider.

As I have said if launch from ssh, it will segfault.
If you have env set, then posix api did as user set.
If you haven't set, posix api will ask, and set default to current active display.

Then to provide code's robustness of NOT properly setting env, casuing segfault issue.

henkwiedig · 2024-11-27T06:50:49Z

setenv("DISPLAY", ":0", 0); does what you want, without the if (getenv("DISPLAY") == NULL) ...

lida2003 · 2024-11-27T06:55:09Z

setenv("DISPLAY", ":0", 0); does what you want, without the if (getenv("DISPLAY") == NULL) ...

This is what we mean: if we don't have this env, it will segfault. If there is no display set, then we simply use default display.
It's a straight forward way to do the default setting. Of course, user can set other values, then there will NOT be NULL.

msposd/osd/util/Render_x86.c

Lines 79 to 85 in cabc6b8

    
                   if (getenv("DISPLAY") == NULL) { 
        
                       /* 
        
                        * Use default display screen, especially launch from console 
        
                        * ToDo: x86 linux should implement this code also, test needed. 
        
                        */ 
        
                       setenv("DISPLAY", ":0", 1); 
        
                   }

henkwiedig · 2024-11-27T07:06:05Z

this

setenv("DISPLAY", ":0", 0);

is equivalant to this:

         if (getenv("DISPLAY") == NULL) { 
             /* 
              * Use default display screen, especially launch from console 
              * ToDo: x86 linux should implement this code also, test needed. 
              */ 
             setenv("DISPLAY", ":0", 1); 
         }

hence the "0". from man 3 setenv

.... If name does exist in the environment, then its value is changed to value if overwrite is nonzero; if overwrite is zero, then the value of name is not changed ....

Edit: we can remove the ifdef in my opinion

lida2003 · 2024-11-27T07:08:51Z

You can do the test. If you don't have this code. It'll be a segfault launching from ssh.

EDIT: Or did you mean use setenv("DISPLAY", ":0", 0); ?

henkwiedig · 2024-11-27T07:16:30Z

Yes just use the non overwrite option from setenv. "0" as thid argument. Then the whole "if" part is useless.
We can also remove the ifdef.

All boils down to one line ! More readable ;)

lida2003 · 2024-11-27T07:39:22Z

@henkwiedig It doesn't work on jetson.

Can you check if the below is what we meant to do.

If I didn't mis-understand, please let me know about setenv

I think lanuching from remote ssh(putty) mgiht be different or something about the understanding of the posix API.

daniel@daniel-nvidia:~/Work/msposd/release/jetson$ echo $DISPLAY

daniel@daniel-nvidia:~/Work/msposd/release/jetson$ ./msposd --master 127.0.0.1:14560  --osd -r 50 --ahi 1 --matrix 11 -v
Ver: 8a82c65-dirty Compiled at: 20241127_153132
Listen on port :127.0.0.1:14560
MSP to OSD mode!Verbose mode!Majestic width:1920,height:1080
Cannot open display
Loading /home/daniel/Work/msposd/release/jetson/font.png for 1000p mode
Font file res 36:13824 pages:1
Loading small size glyphs /home/daniel/Work/msposd/release/jetson/font_hd.png for 1000p mode
Glyph size:36:54 on a 53:20 matrix. Overlay 1912:1080
Create_region PixelFormat:11 Size: 1912:1080 X_Offset:8 Y_Offset:0 results: -1
Segmentation fault (core dumped)
daniel@daniel-nvidia:~/Work/msposd/release/jetson$ git diff
diff --git a/osd/util/Render_x86.c b/osd/util/Render_x86.c
index a49942b..9f42317 100644
--- a/osd/util/Render_x86.c
+++ b/osd/util/Render_x86.c
@@ -72,20 +72,8 @@ int Init_x86(uint16_t *width, uint16_t *height) {
             forcefullscreen=false;
 #ifdef _DEBUG_x86
         forcefullscreen=false;
+        setenv("DISPLAY", ":0", 0);
 #endif
-
-
-#ifdef _x86
-        if (getenv("DISPLAY") == NULL) {
-            /*
-             * Use default display screen, especially launch from console
-             * ToDo: x86 linux should implement this code also, test needed.
-             */
-            setenv("DISPLAY", ":0", 1);
-        }
-#endif
-
-

         display = XOpenDisplay(NULL);
         if (!display) {

henkwiedig · 2024-11-27T07:43:47Z

I assume _DEBUG_x86 was set durting compile time ?

lida2003 · 2024-11-27T07:51:24Z

_DEBUG_x86 was set durting compile time ?

YES

EDIT: check PR-diff

lida2003 · 2024-11-27T07:55:43Z

@henkwiedig
Ah..... it's _DEBUG_x86 not _x86

what this MACRO for?

lida2003 · 2024-11-27T08:00:17Z

@henkwiedig I'll left _DEBUG_x86 as it is. Just optimize setenv for native build.

BTW, it works on jetson, should work on x86 either, as posix api.

EDIT: add _DEBUG_x86 MACRO optimization & clarification #33 for _DEBUG_x86 clarification.

henkwiedig · 2024-11-27T16:40:23Z

Can you rabse this on main. It conflicts with the version.h PR.

@tipoman9 If you don't veto I would merge this.

tipoman9 · 2024-11-27T18:34:04Z

@henkwiedig Sure, go ahead.

Discussion: [ Ignore msposd_jetson for gs on jetson orin OpenIPC#29 ](OpenIPC#29)

lida2003 · 2024-11-28T00:43:22Z

@henkwiedig rebased and worked on jetson.

BTW, please close #29 when it's OK, thanks.

daniel@daniel-nvidia:~/Work/msposd$ file msposd
msposd: ELF 64-bit LSB shared object, ARM aarch64, version 1 (SYSV), dynamically linked, interpreter /lib/ld-linux-aarch64.so.1, BuildID[sha1]=a3fbab226e66a20534c2a495de18003b63d52610, for GNU/Linux 3.7.0, with debug_info, not stripped
daniel@daniel-nvidia:~/Work/msposd$ cat version.h
#ifndef VERSION_H
#define VERSION_H
#define GIT_VERSION "a4ff7f6"
#endif // VERSION_H
daniel@daniel-nvidia:~/Work/msposd$ git log -n 1
commit a4ff7f61675b78f06289a29eb5e1ec31ba552824 (HEAD -> pr_use_native_build, origin/pr_use_native_build)
Author: Daniel Li <lida_mail@163.com>
Date:   Wed Nov 27 15:59:10 2024 +0800

    Optimize setenv for native build
daniel@daniel-nvidia:~/Work/msposd$ cp msposd ./release/jetson/msposd
daniel@daniel-nvidia:~/Work/msposd$ cd ./release/jetson/
daniel@daniel-nvidia:~/Work/msposd/release/jetson$ ./msposd --master 127.0.0.1:14560  --osd -r 50 --ahi 1 --matrix 11 -v
Ver: a4ff7f6 Compiled at: 20241128_083816
Listen on port :127.0.0.1:14560
MSP to OSD mode!Verbose mode!Majestic width:1920,height:1080
Loading /home/daniel/Work/msposd/release/jetson/font.png for 1080p mode
I4 40:13824 Stride:20
Font file res 40:13824 pages:1
Loading small size glyphs /home/daniel/Work/msposd/release/jetson/font_hd.png for 1080p mode
I4 24:9216 Stride:12
Glyph size:36:54 on a 53:20 matrix. Overlay 1920:1080
Glyph size:36:54 on a 53:20 matrix. Overlay 1920:1080
Create_region PixelFormat:3 Size: 1920:1080 X_Offset:0 Y_Offset:0 results: -1
set_LOGO with u32Height:1080 enPixelFormat 3
Listening UDP on 127.0.0.1:14560...

lida2003 mentioned this pull request Nov 25, 2024

Ignore msposd_jetson for gs on jetson orin #29

Closed

lida2003 force-pushed the pr_use_native_build branch from 5f539b2 to 31e8542 Compare November 26, 2024 00:00

lida2003 added 3 commits November 28, 2024 08:34

Use native build instead of x86/jetson target

69b6900

Discussion: [ Ignore msposd_jetson for gs on jetson orin OpenIPC#29 ](OpenIPC#29)

Fix workflow-build for adding native build

58a07d6

Optimize setenv for native build

a4ff7f6

lida2003 force-pushed the pr_use_native_build branch from f7b4ee2 to a4ff7f6 Compare November 28, 2024 00:36

henkwiedig merged commit 60b2990 into OpenIPC:main Nov 28, 2024
1 of 9 checks passed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Use native build instead of x86/jetson target #31

Use native build instead of x86/jetson target #31

lida2003 commented Nov 25, 2024

lida2003 commented Nov 25, 2024

henkwiedig commented Nov 26, 2024

lida2003 commented Nov 26, 2024

henkwiedig commented Nov 26, 2024

lida2003 commented Nov 26, 2024 •

edited

Loading

henkwiedig commented Nov 27, 2024

lida2003 commented Nov 27, 2024

henkwiedig commented Nov 27, 2024

lida2003 commented Nov 27, 2024 •

edited

Loading

henkwiedig commented Nov 27, 2024 •

edited

Loading

lida2003 commented Nov 27, 2024 •

edited

Loading

henkwiedig commented Nov 27, 2024 •

edited

Loading

lida2003 commented Nov 27, 2024

henkwiedig commented Nov 27, 2024

lida2003 commented Nov 27, 2024 •

edited

Loading

lida2003 commented Nov 27, 2024

lida2003 commented Nov 27, 2024 •

edited

Loading

henkwiedig commented Nov 27, 2024 •

edited

Loading

tipoman9 commented Nov 27, 2024

lida2003 commented Nov 28, 2024 •

edited

Loading

Use native build instead of x86/jetson target #31

Use native build instead of x86/jetson target #31

Conversation

lida2003 commented Nov 25, 2024

lida2003 commented Nov 25, 2024

henkwiedig commented Nov 26, 2024

lida2003 commented Nov 26, 2024

henkwiedig commented Nov 26, 2024

lida2003 commented Nov 26, 2024 • edited Loading

henkwiedig commented Nov 27, 2024

lida2003 commented Nov 27, 2024

henkwiedig commented Nov 27, 2024

lida2003 commented Nov 27, 2024 • edited Loading

henkwiedig commented Nov 27, 2024 • edited Loading

lida2003 commented Nov 27, 2024 • edited Loading

henkwiedig commented Nov 27, 2024 • edited Loading

lida2003 commented Nov 27, 2024

henkwiedig commented Nov 27, 2024

lida2003 commented Nov 27, 2024 • edited Loading

lida2003 commented Nov 27, 2024

lida2003 commented Nov 27, 2024 • edited Loading

henkwiedig commented Nov 27, 2024 • edited Loading

tipoman9 commented Nov 27, 2024

lida2003 commented Nov 28, 2024 • edited Loading

lida2003 commented Nov 26, 2024 •

edited

Loading

lida2003 commented Nov 27, 2024 •

edited

Loading

henkwiedig commented Nov 27, 2024 •

edited

Loading

lida2003 commented Nov 27, 2024 •

edited

Loading

henkwiedig commented Nov 27, 2024 •

edited

Loading

lida2003 commented Nov 27, 2024 •

edited

Loading

lida2003 commented Nov 27, 2024 •

edited

Loading

henkwiedig commented Nov 27, 2024 •

edited

Loading

lida2003 commented Nov 28, 2024 •

edited

Loading