The issue of the day: Inferno panic.
Ok, I’ve managed to do the thing with arrays in Inferno C Loader module. The case was that I should have used Array of Pointers to create Array of Arrays. Thanks Powerman from Inferno’s IRC! The new code:
ar = H2D(Array*, heaparray(&Tptr, nimports));
ai = (Array**)ar->data;
for(i2 = m->ldt; *i2 != nil; i2++){
nimports = 0;
for(i1 = *i2; i1->name != nil; i1++)
nimports++;
*ai = H2D(Array*, heaparray(Timport, nimports));
li = (Loader_Import*)(*ai)->data;
for(i1 = *i2; i1->name != nil; i1++){
li->name = c2string(i1->name, strlen(i1->name));
li->sig = i1->sig;
li++;
}
ai++;
} I've completed my changes to the Loader module, now it supports getting and restoring all the Imports. I've also fixed classloader code to not forget this section. External modules load ok now.
So, I’ve returned to my previous bug of Inferno panic. This are the ending lines of java start log:
... 1738 lines of the log skipped ...
modpatch java/lang/Object
compile java/lang/Object
compile native java/lang/Object
Switching class java/lang/Object state to INITED
java/lang/Object->init()
done
java/lang/Object-><clinit>()
clinit done
Switching class java/lang/Class state to INITED
java/lang/Class->init()
done
java/lang/Class-><clinit>()
Debug pring: registerNatives called
clinit done
Switching class java/lang/String state to INITED
java/lang/String->init()
done
java/lang/String-><clinit>()
alloc:D2B: 68a874 in heap: in block at 68a440'640
area at 68a864: ffffffff 437978 43c878 1ac ffffffff 0 ffffffff
68a880: 0 ffffffff 437c78 ffffffff 1cc ffffffff 437cd8 43c8d8
68a8a0: 1dc ffffffff 0 ffffffff 437eb8
panic: alloc:D2B 68a874 (from b0102e28/0)
I need to find out what causes this. It had already happened before in the Class class, and now it also touches String. I know that the error is somewhere in the class static init function, and from the Class case analythis it seems that array initalisation has something to do with this bug.
Happily, Charles suggested to use gdb (and I thought that Inferno is compiled using its own compilers so no debbugging info), and that worked! I’ve fixed makefiles (added -g to the gcc), and now I can debug that awful bug! By the way, Inferno (and currently libinterp) sources are really quite clear and easy to understand, I really love them!
To be continued…
Worked on files:
- libinterp/loader.c: Loader module source code
- modules/Loader.m: module definition file
- java/pkg/java/lang/String_L.[bm]: String_L native class