j.isorce, can you start looking at how your previous CL could be fixed to make the unittests pass on Linux with "AMD" or "ATI" as driver vendor for the last entry? You can use "git cl patch https://codereview.chromium.org/2198923002".

Sure I will have look (I do not get notifications on this email at home)

I just found a fix, I will merge into the original CL once the revert has landed:

--- a/gpu/config/gpu_info_collector_linux.cc
+++ b/gpu/config/gpu_info_collector_linux.cc
@@ -247,6 +247,9 @@ CollectInfoResult CollectBasicGraphicsInfo(GPUInfo* gpu_info) {
 CollectInfoResult CollectDriverInfoGL(GPUInfo* gpu_info) {
   DCHECK(gpu_info);
 
+  if (!gpu_info->driver_vendor.empty() && !gpu_info->driver_version.empty())
+    return kCollectInfoSuccess;
+

Indeed there are some specific functions to retrieve driver vendor and version, like CollectDriverVersionATI and CollectDriverVersionNVidia. The extraction from the GL_VERSION should only be done when specific functions cannot be used.

Revert landed, you can go ahead.

Done in "Patch Set 12" of https://codereview.chromium.org/2153373002/

Please make sure cwallez@'s new unit test has landed before attempting to reland this patch.

I didn't realize that the unit test in https://codereview.chromium.org/2198923002/ has been abandoned. Please figure out a way to better test https://codereview.chromium.org/2153373002/ to prevent this regression from happening in the future.

Oki I will try to make a test for that error. In the meantime I have submitted "Patch Set 13": https://codereview.chromium.org/2153373002/

Just for the log I have integrated that test (#c6) in the CL since then. The CL https://codereview.chromium.org/2153373002/ has landed so marking the issue as fixed. I keep https://bugs.chromium.org/p/chromium/issues/detail?id=628929 open to track any regression.