Building enso from source fails at kernal.c

[veerpandya]

I'm trying to build henkaku enso from source and running make keeps giving me this error:

[  4%] Building C object CMakeFiles/emmc_helper.dir/src/kernel.c.obj
In file included from /Users/kou/Downloads/enso-1.1/installer/src/kernel.c:10:
/usr/local/vitasdk/arm-vita-eabi/include/libk/stdarg.h:17:25: error: conflicting types for 'va_list'
   17 | typedef _PDCLIB_va_list va_list;
      |                         ^~~~~~~
In file included from /usr/local/vitasdk/arm-vita-eabi/include/psp2kern/kernel/debug.h:12,
                 from /usr/local/vitasdk/arm-vita-eabi/include/psp2kern/kernel/sysmem.h:10,
                 from /Users/kou/Downloads/enso-1.1/installer/src/kernel.c:5:
/usr/local/vitasdk/lib/gcc/arm-vita-eabi/10.1.0/include/stdarg.h:99:24: note: previous declaration of 'va_list' was here
   99 | typedef __gnuc_va_list va_list;
      |                        ^~~~~~~
make[2]: *** [CMakeFiles/emmc_helper.dir/src/kernel.c.obj] Error 1
make[1]: *** [CMakeFiles/emmc_helper.dir/all] Error 2
make: *** [all] Error 2

I'm not sure if this is the right place to post this issue, but it seems to be some kind of naming issue with the va_list in the different kernal.c files. Any help would be greatly appreciated!

[Princess-of-Sleeping]

Currently confirming this issue.

[Princess-of-Sleeping]

If you want to fix it temporarily, do as follows

first, do vitasdk-update

vvv second, add patch vvv
kernel.c line 10

+ #include <psp2kern/kernel/sysclib.h>
+ /*
  #include <libk/stdarg.h>
  #include <libk/string.h>
  #include <libk/stdio.h>
+ */

main.c line 12

  #include <psp2/io/stat.h>
+ #include <psp2/update.h>
  #include <taihen.h>

main.c line 26

  int _vshSblAimgrGetConsoleId(char cid[32]);
- int sceSblSsUpdateMgrSetBootMode(int x);
  int vshPowerRequestColdReset(void);

[veerpandya]

This fixed it, thanks so much!

[veerpandya]

Oh actually, so the build went through fine, but when I tried to install enso from the installer on my vita it gave me this error:

Failed to load kernel workaround:  0x80010002

Also I should add, the logo wasn't showing on the enso app in my Vita, it was just a white bubble. Not sure if this is a related issue.

[veerpandya]

I've unzipped the enso_installer.vpk and it's only showing eboot.bin and the sce_sys folder. The released version of enso has a kernel2.skprx, along with two emmc_helper files and the fat.bin. My installer is also missing the icon.png, bg.png, and template.xml files, which explains why the app didn't show the logo on my vita. I'm guessing this is an issue with my make or cmake command? I didn't get any errors when running them and I'm using cmake v2.8.12.2. I'm fairly new to vita development so it's very possible that I'm overlooking something.

[veerpandya]

Ok I just manually added the missing files from the build folder to the new enso_installer.vpk and it worked perfectly. Not sure why make didn't add it, but the issue I had originally is definitely fixed. Thanks again!

[frangarcj]

Tracked this down for real. libk's _PDCLIB_va_list is char *, and it walks it as a flat sequence of stack slots — that's what conflicts with the compiler's own va_list the moment a file includes both libk/stdarg.h and anything that needs sceClibVprintf (which is exactly this combination: psp2kern/kernel/debug.h pulls in the real one, libk/stdarg.h redeclares it). Confirmed it still reproduces today, unchanged, against the current SDK.

There's actually a second problem hiding behind it: even without the redeclaration clash, that char* walk silently returns wrong values for long long/double/anything 8-byte once you're past the register-passed arguments — AAPCS aligns those to an even register pair and the macro doesn't account for it. int/long/pointers come back fine, which is why this never showed up from the arithmetic side before.

Opened a fix upstream: DaveeFTW/libk#5 — swaps the char* hack for the compiler's own va_list/va_arg.

If you're hitting a different, earlier error first — Unsupported __STDC_VERSION__ — that's unrelated (libk's headers only accept exactly C99, and current gcc defaults to newer), and adding -std=gnu99 to your build flags works around it on your own end. That flag alone does not fix what this issue is about though — the va_list conflict still happens either way until the fix above lands.

Leaving this open until vitasdk's own libk package actually carries the fix — it doesn't yet.