Giters

hanatos / vkdt

raw photography workflow that sucks less

Home Page:https://vkdt.org

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

vkdt crashes when opening with image location on command line

ptilopteri opened this issue · comments

commented

vkdt ./.
will crash

./. contains a few jpgs and some nefs

backtrace attached
vkdt-bt-15817.txt

openSUSE Tumbleweed 20231028
Linux 6.5.9-1-default x86_64
20-thr 12th Gen Intel(R) Core(TM) i7-12700 4900MHz
Load: 0.29 Tasks: 1427
Mem: 6162/31807MB
Disk: 6956/15831GB
Gfx: @ 5360x1440

commented

hm can of course not reproduce just like that.

did you use rawspeed or rawler as a backend for raw loading?

also if you make debug instead of make it'll include the debugging symbols. all i can tell from the stripped stack trace is that it's going wrong somewhere in the i-raw module.

did it say something on the console when it died? the rust backend usually drops a helpful line or two there..

commented

hm can of course not reproduce just like that.

did you use rawspeed or rawler as a backend for raw loading?

I only ran: vkdt ./.
to load files from a local directory

also if you make debug instead of make it'll include the debugging symbols. all i can tell from the stripped stack trace is that it's going wrong somewhere in the i-raw module.

I ran "make debug j12"

did it say something on the console when it died? the rust backend usually drops a helpful line or two there..

[Thread debugging using libthread_db enabled]
Using host libthread_db library "/lib64/libthread_db.so.1".
0x00007f4621700b17 in wait4 () from /lib64/libc.so.6
backtrace written to /tmp/vkdt-bt-25789.txt
recovery data written to /tmp/vkdt-crash-recovery.*

vkdt-bt-25789.txt
backtrace after "make deb.."

if I run "vkdt ./*.jpg" it doesn't crash but only loads one file
I can then use "escape" -> collect
and load the intended files

tks

commented

hm still no debug symbols in this one. at least it seems to be reproducible.

did you make clean before the debug build? i.e.

make clean && make debug -j20

if you didn't put a custom bin/config.mk into place (starting from comments/values in bin/config.mk.defaults) you'll be running without exiv2 and with rawspeed. the stack trace seems a few layers deeper than my last function (so inside rawspeed?) which is weird because you say opening the folder from the gui works?

commented

I have added config.mk to include exiv2
updated system, new kernel and revised nvidia video driver

now running vkdt alone on the cl provides:

vkdt
[gui] glfwGetVersionString() : 3.3.8 X11 GLX EGL OSMesa clock_gettime evdev shared
[gui] monitor [0] HDMI-0 at 3440 0
[gui] monitor [1] DP-1 at 0 0
[gui] vk extension required by GLFW:
[gui] VK_KHR_surface
[gui] VK_KHR_xcb_surface
[ERR] could not load function vkDebugMarkerSetObjectNameEXT. do you have validation layers setup correctly?
[ERR] init vulkan failed
[ERR] failed to init gui/swapchain

note: vkdt will open if I do not make debug j20
but will only open jpg files, not nef's
vkdt crashes on trying to open a nef

commented

make debug j20
make.txt

commented

ran make w/o debug
vkdt will open and edit jpg's but not nef raws
attempting to open a nef crashes
attach backtrace but no debug data
vkdt-bt-14191.txt

commented
commented
commented
commented
commented

ok, ran:
make clean
make debug -j12
output: make.txt

ran ./bin/vkdt

[gui] glfwGetVersionString() : 3.3.8 X11 GLX EGL OSMesa clock_gettime evdev shared
[gui] monitor [0] HDMI-0 at 3440 0
[gui] monitor [1] DP-1 at 0 0
[gui] vk extension required by GLFW:
[gui] VK_KHR_surface
[gui] VK_KHR_xcb_surface
[ERR] error VK_ERROR_LAYER_NOT_PRESENT executing vkCreateInstance!
[ERR] did you install the vulkan validation layer package?
[ERR] init vulkan failed
[ERR] failed to init gui/swapchain

vulkan-validationlayers IS installed

ran ./bin/vkdt /data1/testing/*.jpg

[gui] glfwGetVersionString() : 3.3.8 X11 GLX EGL OSMesa clock_gettime evdev shared
[gui] monitor [0] HDMI-0 at 3440 0
[gui] monitor [1] DP-1 at 0 0
[gui] vk extension required by GLFW:
[gui] VK_KHR_surface
[gui] VK_KHR_xcb_surface
[ERR] error VK_ERROR_LAYER_NOT_PRESENT executing vkCreateInstance!
[ERR] did you install the vulkan validation layer package?
[ERR] init vulkan failed
[ERR] failed to init gui/swapchain

openSUSE Tumbleweed 20231030
Linux 6.5.9-1-default x86_64
20-thr 12th Gen Intel(R) Core(TM) i7-12700 4900MHz
Load: 0.13 Tasks: 1569
Mem: 7030/31807MB
Disk: 6973/15831GB
Gfx: @ 5360x1440
NVIDIA GA104 [GeForce RTX 3060 Ti Lite
nvidia v: 535.113.01

X11

no crash logs generated

commented

ran:
make clean
make

./bin/vkdt
vkdt opens to lightable displaying previous images, jpg and nef

can open jpg to edit
crashes when opening nef
bt: vkdt-bt-10695.txt

ps: does not exit vkdt

commented

wiped vkdt git
cd ./vkdt
did git clone https://github.com/hanatos/vkdt
git submodule init; git supmodule update
cd ./bin
make clean; make

now cannot open vkde with or without bare commandline
no crash

[gui] glfwGetVersionString() : 3.3.8 X11 GLX EGL OSMesa clock_gettime evdev shared
[gui] monitor [0] HDMI-0 at 3440 0
[gui] monitor [1] DP-1 at 0 0
[gui] vk extension required by GLFW:
[gui] VK_KHR_surface
[gui] VK_KHR_xcb_surface
[ERR] error VK_ERROR_LAYER_NOT_PRESENT executing vkCreateInstance!
[ERR] did you install the vulkan validation layer package?
[ERR] init vulkan failed
[ERR] failed to init gui/swapchain

tks

commented

hm all that sounds very incoherent to me. so at some point it did open nef but now doesn't? or just shows them as thumbnails (which goes through the same code path) but doesn't open in darkroom mode? also now it also has validation layer problems even if not built in debug mode (where it clearly doesn't use validation layers ..)?

i also tried to use the rawspeed/exiv2 codepath locally and works here. did you make install at some point and now have a mixture of old and new binaries?

maybe you can share one .nef file that reproduces the issue reliably? maybe there's something particular about the exif data? running out of ideas here.

commented

i also tried to use the rawspeed/exiv2 codepath locally and works here. did you make install at some point and now have a mixture of old and new binaries?

no, I never use "make install"

maybe you can share one .nef file that reproduces the issue reliably? maybe there's something particular about the exif data? running out of ideas here.

attached subj nef
nef to too large, see http://wahoo.no-ip.org/~paka/231026_125122_aab_1139.nef

I will wipe clean and clone anew and report.

tks for attention and sorry for bother

commented

rm -rf ./vkdt/. ./vkdt/.*
git clone https://github.com/hanatos/vkdt
git submodule init; git submodule update
make debug -j12

./vkdt/bin/vkdt
[gui] glfwGetVersionString() : 3.3.8 X11 GLX EGL OSMesa clock_gettime evdev shared
[gui] monitor [0] DVI-I-3 at 2560 0
[gui] monitor [1] HDMI-0 at 0 0
[ERR] couldn't get GLFW instance extensions
[ERR] failed to init gui/swapchain

rpm -qa |grep -i glfw
libglfw3-3.3.8-2.3.x86_64
libglfw-devel-3.3.8-2.3.x86_64

does this help?
tks

commented

strike the output from ./vkdt/bin/vkdt

should be:

[gui] glfwGetVersionString() : 3.3.8 X11 GLX EGL OSMesa clock_gettime evdev shared
[gui] monitor [0] HDMI-0 at 3440 0
[gui] monitor [1] DP-1 at 0 0
[gui] vk extension required by GLFW:
[gui] VK_KHR_surface
[gui] VK_KHR_xcb_surface
[ERR] init vulkan failed
[ERR] failed to init gui/swapchain

tks

commented

I will close. we can attribute my problems to nvidia drivers, I think.

a driver update and a subsequent system reboot provide a working vkdt.

sorry for the bother and thanks for the attention.