Be more lenient with shader matching.

[apitrace-tests] / trim_stress / README.markdown

diff --git a/trim_stress/README.markdown b/trim_stress/README.markdown
index adab8dcb1b74117d924ba2e7d004633a8a6f2385..ed872fc08e0f161bd4d55b29970ea944ee4117c8 100644 (file)
--- a/trim_stress/README.markdown
+++ b/trim_stress/README.markdown
@@ -5,15 +5,28 @@ will perform the following operations:
 
 Given <program>.trace:
 
-    1. Generate snapshots of original trace in ./<program>-ref/
+    1. Generate snapshots of original trace in ./<program>-ref/, (this
+       step is skipped if the <program>-ref directory already exists.
+       Simply delete this directory before "make test" to force new
+       generation of reference images.
 
     2. For each frame of the trace:
 
         1. Use "apitrace trim" to trim <program>.trace to a
            single-frame trace in <program>-trim.trace
 
-       2. Generate a snapshot of the trimmed trace to
-          ./<program>-out/<frame>.png
+       2. Generate a snapshot of the trimmed frame to
+          ./<program>-tmp-out/<frame>.png
 
-    3. Use "apitrace diff-images" to compare all snapshots in
-       ./<program>-out with those in ./<program>-ref
+        3. Use "apitrace diff-images" to compare this one frame with
+           the corresponding reference image in ./<program>-ref.  If
+           the images differ the entire process is interrupted here.
+
+        4. Move the frame snapshot from ./<program>-tmp-out to
+           ./<program>-out
+
+Note that you can also put traces in the sub-directory local and they
+will be tested as well. The difference is that those files will be
+ignored by git. This can be useful for traces that are too large to be
+considered for inclusion for git, or for traces that cannot be shared
+due to proprietary contents.