flypig.co.uk

I've been really struggling to get the GLScreenBuffer code back compiling again over the last few days. This is partly because a whole raft of interfaces have changed and these changes have seeped through into many different places in the code. Often the changes are quite small, such as using a reference rather than a pointer. But other times they're more significant, such as the way member variables in SharedSurface have changed. To look into the latter in a little more detail, the previous version looked like this:

class SharedSurface {
 public:
[...]
  const SharedSurfaceType mType;Th
  const AttachmentType mAttachType;
  const WeakPtr<GLContext> mGL;
  const gfx::IntSize mSize;
  const bool mHasAlpha;
  const bool mCanRecycle;

But the new code bundles all this up into a structure so that it now looks more like this:

struct PartialSharedSurfaceDesc {
  const WeakPtr<GLContext> gl;
  const SharedSurfaceType type;
  const layers::TextureType consumerType;
  const bool canRecycle;
};
struct SharedSurfaceDesc : public PartialSharedSurfaceDesc {
  gfx::IntSize size = {};
};

class SharedSurface {
 public:
[...]
  const SharedSurfaceDesc mDesc;

Even this might seem like a small change, but it can make it really hard to match up method signatures given the ESR 78 version takes a collection of individual parameters, whilst the ESR 91 version requires just a single SharedSurfaceDesc instance.

Besides all this I've also had what feels like several long days at work. The result is that when I come to work on Gecko of an evening the code swims around in front of my eyes and refuses to stay still as my mind drifts hither and thither.

Nevertheless I've been persevering with my "fix, compile, examine" routine and have continued to make some slight progress. Once again, I don't want to go into the full detail of the changes I've had to make, because it really is just reintroducing code that was removed upstream. So there's a lot of it, and I'm not even spending any time attempting to understand it.

But it has now got to the point where all three directories I've been making changes to are compiling when I do the partial builds:

$ make -j1 -C obj-build-mer-qt-xr/gfx/gl/
[...]
$ make -j1 -C obj-build-mer-qt-xr/dom/canvas
[...]
$ make -j1 -C obj-build-mer-qt-xr/gfx/layers
[...]

I've been here before of course. Just because the partial builds compile without errors doesn't mean the full build will pass. So the next step for me tonight is to set the full build running so we can come back to it and check its status in the morning.

Before I do that, here are the latest stats showing the changes I've made in relation to the offscreen rendering pipeline:

$ git diff --numstat
73      16      gfx/gl/GLContext.cpp
73      8       gfx/gl/GLContext.h
11      0       gfx/gl/GLContextTypes.h
590     0       gfx/gl/GLScreenBuffer.cpp
171     1       gfx/gl/GLScreenBuffer.h
305     5       gfx/gl/SharedSurface.cpp
132     4       gfx/gl/SharedSurface.h
16      10      gfx/gl/SharedSurfaceEGL.cpp
13      3       gfx/gl/SharedSurfaceEGL.h
81      2       gfx/gl/SharedSurfaceGL.cpp
61      0       gfx/gl/SharedSurfaceGL.h
60      0       gfx/gl/SurfaceTypes.h
21      0       gfx/layers/client/TextureClientSharedSurface.cpp
25      0       gfx/layers/client/TextureClientSharedSurface.h
$ git diff --shortstat
 14 files changed, 1632 insertions(+), 49 deletions(-)

That's not a huge increase on yesterday, but is at least non-zero. This reflects the fact that as things have progressed the fixes have become smaller, but thornier. All of the main changes this time seem to have been to the GLContext code.

Alright, time to set the build off. Hopefully this will complete without errors, but I'm not betting on it.

If you'd like to read any of my other gecko diary entries, they're all available on my Gecko-dev Diary page.