logs archiveBotHelp.net / Freenode / #3dsdev / 2015 / August / 24 / 1
endrift
also does anyone know the maximum binary size for ironhax?
Subv
it transfers data from the linear heap to vram or vice-versa, and lets do perform some transformations on the data while it transfers
endrift
is this a function call?
Is there more information available?
Subv
http://3dbrew.org/wiki/GSP_Shared_Memory#Trigger_Display_Transfer
see http://3dbrew.org/wiki/GPU/External_Registers#Transfer_Engine for a list of flags
you can make it tile / untile your data during the transfer, you can make it apply a box filter, you can make it flip your data vertically
xerpi
wow
I should use this
endrift
looks like there's no associated API call yet?
yuriks
endrift: GX_SetDisplayTransfer
endrift
oh, GX
okay
I'm running up against the maximum binary size at the moment, it seems :(
wait CitrAGB is much larger
? ? ?
yuriks
hmmm
you also have limits for individual sections
like, your .text (code) has to fit in whatever executable memory the exploit has available for use
endrift
http://pastie.org/10371257
yuriks
I don't know the limits, smea would know
and you should ask shinyquasire what's the .text size for CitrAGB when he comes on
endrift
I could try to compile it myself
but that would require compiling it myself
yuriks
haha
theoretically you could extract it from the 3dsx header
endrift
oh
wow it's totally just there
1109780 bytes
larger than my binary is entirely
I wonder if the binary's not getting relocated properly
should I just bug smea and toss him a binary?
yuriks
*shrug* I guess
endrift
wait
it's totally working now
????
Subv
welcome to 3DS homebrew development
just wait until the GPU decides to crash on you for no goddamn reason, then your initiation will be finished
endrift
My kingdom for an exception handler
yuriks
endrift: yay!?:D
endrift
I guess!
now I can debug real bugs instead of fake bugs!
yuriks
meanwhile I'm banging at sign until I manage to convince the GPU to show something on the screen
signs*
endrift
now it's doing the inexplicable crash thing again
ugh I guess I'm done for now
yuriks
~3DS dev~
Subv
endrift: your initiation is now complete
Lectem
endrift : sounds like some stack corruption or bad ptr
I had random crashes in an app before because of this
give citra a shot, it could help
witheld
StapleButter: wait, what other channel is there? This is the channel on the wiki
StapleButter
irc.efnet.net, #3dsdev
most of the elite is there
Subv
elite, heh
witheld
wait this isn't the channel on the wiki
how did I get here???
endrift
hah
do breakpoints even work in Citra?
citra-qt seems to be very broken
Subv
pica breakpoints work. anything cpu-debug related is broken
endrift
G R E A T
well it's not crashing in Citra, regardless
it's clearly an issue on hardware only
Subv
you could debug using printf
witheld
Just debug citra to debug your program!
/s
Subv
^ i've actually done that several times to debug problematic games in citra
endrift
yeah I've tried that
I mean I know where it's crashing
it's on an assignment from memory that exists to memory that exists
so it makes no sense
witheld
Well, it follows that clearly you should be making assignments from memory that doesn't exist to memory that doesn't exist
Since that oposite doesn't work
Subv
endrift: are you messing with the linear heap and/or vram?
the cpu can't write to vram directly, and the gpu can't write to the normal heap directly
endrift
VRAM, no
Subv
endrift: is your code online?
endrift
not latest
Subv
can you show us a snippet of the code that's crashing?
Lectem
endrift : perhaps your call depth is too high? the stack isnt big
endrift
hmm, maybe
Lectem
also do you have a lot of objects allocated on the stack?
like big ones
endrift
seems unlikely though
no, all on the heap
the assignment in question is also on the heap
it's .text address -> heap location
yuriks
you can increase the stack size by adding a magic global somewhere:
u32 __stacksize__ = 256 * 1024;
endrift
works in citra ow
now
crashes on hardware
but it looks really cool when it crashes
yuriks
haha
endrift
I gave up on libpng
turns out embedding the font as a pre-tiled bitmap was smaller
than all of libpng
go figure
64kB vs like 300kB
« prev 1 2 3 next »