The next release of #Lagrange arrives as March draws to a close:
There's lots to unpack here: UI localization, LibreTranslate integration, collapsing preformatted blocks, alt text, trust via CAs, theme adjustments, bold link appearance with custom icons, touch screen support, runtime UI rescaling, word/paragraph selection, unread feed entry counter, basic command line use, and several UI improvements.
@jk I don't know if you already knew about this, but one problem I had was keybindings for going back in history and stuff still worked when pressed F to view hints. It's not supposed to.
@jk thanks for the new release. Unfortunately, on OpenBSD Lagrange instantly crashes after displaying the window.
I'll open an issue on the project forge.
@solene Did a clean OpenBSD VM install and built Lagrange. Everything seems to work, though.
Are you able to get a backtrace in a debugger, to see where it's crashing?
@jk did you try with OpenBSD 6.8 ? I'm using the development version -current, to switch to it you can run "sysupgrade -s" as root and will download the -current sets, reboot to install them and do the work.
in one case I can get a backtrace with gdb, in the other case with an empty directory folder I only get a message and no core dump.
@jk when I have no .config/lagrange directory I have this output and no core dump :/
Lagrange: A Beautiful Gemini Client
[window] renderer: opengl (accelerated)
i965: Failed to submit batchbuffer: Bad file descriptor
usually, I don't have the last line.
~/.config/lagrange get created and a window briefly get displayed (but only to see it appear and disappear, I can't see anything it it so it's very short).
@solene > i965: Failed to submit batchbuffer: Bad file descriptor
This suggests to me that there is a problem with Intel graphics drivers. Is that what you're using?
Does it work if you launch with `--sw`?
@jk it works with --sw
I have an intel graphics and 1.2.3 works fine. I get this console output when I start 1.2.3 [window] renderer: opengl (accelerated)
@jk so, 1.3.0 starts with an empty config directory and --sw
meanwhile, if I use my config directory and run lagrange with --sw it works too. So it seems there is only one issue that trigger two behaviors
@solene In that case I probably can’t reproduce this in the VM since it uses its own graphics driver.
My theory is that Lagrange tries to cache glyphs too early (before the window is shown) and the SDL renderer isn’t fully up-and-running yet (?). I can try to defer caching until a bit later.
@jk I restarted my X session, now 1.3.0 starts fine, but I still have the same ""i965: Failed to submit batchbuffer: Bad file descriptor" issue without a ~/.config/lagrange
I hope this helps
I’ll apply a flag that enables text caching only after the window has been exposed, to ensure that graphics can be drawn successfully.
@solene Could you give this a try: https://git.skyjake.fi/skyjake/lagrange/commit/8e8e486a2969dbc20b3c3a08023267e9eab53265
It should prevent graphics from being drawn too early.
@jk I tried your changes by downloading the commit as a patch and applying it. I still have the « i965: Failed to submit batchbuffer: Bad file descriptor » issue with an empty configuration file.
It still work with my configuration directory.
By playing with ~/.config/lagrange/ content I pinpointed something.
mkdir lagrange ; cd lagrange
with that, lagrange starts! Hope this help :)
@solene Interesting clue! This is a bit of a long shot, but what if you change the line window.c:1476 to this:
(I.e., add a ~ in the event ID.)
If that makes no difference, we'll have to somehow pinpoint exactly when the i965 error occurs.
@solene Thanks. I found a failed assertion following your previous steps: https://git.skyjake.fi/skyjake/lagrange/commit/a6e4bceb063af6a3b5a0b4412f39f3e215d5c5ab
Hopefully this will do the trick.
@solene I'll add some extra logging options and we can see if those reveal anything new about the crash.
v1.3.1 is out but I don't expect it to help here...
skyjake's personal Mastodon instance