You are not logged in.

#1 2018-11-20 20:00:15

jamespharvey20
Member
Registered: 2015-06-09
Posts: 129

[SOLVED] Kernel bisection between 4.18 and 4.19 compile error

There's hundreds, maybe around a thousand, of bisection commits that won't compile.  I have no idea how this sort of thing happens.

In short: If you run into this between commits "39b4cbad" (3/30/2018) and "4eaf317a" (7/12/2018), the fix is to "git cherry-pick ef8e0ff9" .  It's a short enough commit, and it seems to be cleanly applying for me so far.

Setting "CONFIG_DRM_I915*" to "=n" does not prevent compilation of this file.  In IRC, I originally thought it did, but was mistaken.

While bisecting an amdgpu bug that prevents booting, between 4.18 and 4.19, I ran into this gem:

drivers/gpu/drm/i915/gvt/kvmgt.c: In function ‘gvt_dma_map_page’:
drivers/gpu/drm/i915/gvt/kvmgt.c:188:17: error: ‘pfn’ undeclared (first use in this function); did you mean ‘gfn’?
  if (!pfn_valid(pfn)) {
                 ^~~
                 gfn
drivers/gpu/drm/i915/gvt/kvmgt.c:188:17: note: each undeclared identifier is reported only once for each function it appears in

OK, sometimes broken source is committed in git trees.  It's too bad it's happened in linux.  Try moving from broken commit "ef8e0ff9" to "ef8e0ff9~3", still fails.  Try moving to ""ef8e0ff9~100", as I still have a long ways to go in the bisection, still fails.  I also ran into this many hundreds of commits away, after my first try of randomly marking this initial failure as good just to proceed.  When I hit this, there's 1167 revisions left, but I'm not sure of the exact number of commits this affects, with how the branches and merges work.

The log for "drivers/gpu/drm/i915/gvt/kvmgt.c" shows commit "4eaf317a", which explains the breaking commit for compilation was "39b4cbad".

I have to figure someone else is going to run into this, so am making the unusual step of making an immediately "[SOLVED]" post.

Offline

Board footer

Powered by FluxBB