Passing through the Fort Pitt tunnel (1:22pm), I received a phone call from an unknown number. seth's new cellphone, actually. He was calling regarding if and when I was coming in today, and was unhappy with my declaration of 30 minutes. Uh oh. But I was already committed to getting my mother to the airport; I violated a few of my driving principles on the return flight, and, despite it being midday traffic on the airport corridor, I was at his office by 1:57pm. Promptly to be told "oh good... I'll come by your office in 5 minutes".
Yeah. Seems he's suddenly got an interest in the what's and how's of the Cmufloat project. The bane of my summer. There is a lot of parts of that system that I was not proud of -- but no one had any interest in using it or how it worked, and it basically got dropped as I was reassigned. But of course, big quarterly (or FINAL?) report... on Friday. So with my ass exposed to all the world -- or at least to his intensive scrutiny, it is going to be a long week.
"I'll be back around 8pm -- will you be here" (not a question, should have been structured "you will be here"). I've spent the afternoon stressed, rewriting and fixing the stuff that had been done. Fought with the open64 compiler madness until I was able to remember the appropriate environment variable trick that made it go.
Only to discover, just now, that the version of gcc (2.96) on the Itanium is now different from that of my machine (2.95.3, the last version that can compile open64). He wants these results tonight, although when he gets back ("I'm running late, I won't be there til 9, is that going to ruin you" (not a question)) we'll have to discuss it.
For those of you with compiler experience, here's the compiler chain: Take a source file. Compile it to ia64 asm on my machine with my modified compiler (open64, uses the gcc frontend). Copy it to the itanium machine. Compile it (with some other object files) to ia64 and run it. Seems to me that having the standard libraries in very different places between the asm file and the machine is basically "game over", right? In hind sight, this was a bad idea, considering that we've mostly just made modifications to the gcc frontend part.
At least I now have corvisdog at my feet, ready to shred seth on command.