tag:blogger.com,1999:blog-6628985022531866193.post1113335707348223768..comments2024-02-12T17:37:05.629+00:00Comments on The OldWood Thing: A Not So Minor Hardware RevisionChris Oldwoodhttp://www.blogger.com/profile/18183909440298909448noreply@blogger.comBlogger1125tag:blogger.com,1999:blog-6628985022531866193.post-71035075053943195882019-03-07T22:00:27.364+00:002019-03-07T22:00:27.364+00:00Reading this I got an increasing sense of deja vu,...Reading this I got an increasing sense of deja vu, but mine was not quite the same. Back when I was working on Railtrack (or software was only a small part of that big disaster) I to got a report from Test that I couldn't reproduce. Again, testers could see it but not developers just like your example.<br /><br />Eventually I got it down to a Visual C++ 2.0 bug which only effected Intel machines. The testers had Intel machines, developers had AMD based Dell's. The compiler was making an optimisation in integer initialization which was fine on AND but failed in Intel. The fix was to disable optimisation.<br /><br />Yes I know we expect it to be the other way around - AND is a copy of Intel after all - but that's why it sticks in my mind.<br /><br />Otherwise, my story is pretty much yours. <br />allan kellyhttps://www.blogger.com/profile/06262139490250478379noreply@blogger.com