02-28-2009 06:41 PM
Just downloaded webPack 10.1 and just to make sure everything is worjing I opened a project that synthesised/implemented just fine on another machine last week. I hit "synthesize" in the "processes" tab and I get the usual list of various informational messages -- no warnings, no errors. At the end it gives me the final timing analysis and then the single line
Process "Synthesis" failed
That's all. In the "processes" tab I get the round red icon with the white cross that would indicate an error, but both the error and the warning tab down at the console are empty. When I click on "design summary" the three topmost lines in the upper right corner say
Current state: synthesized
Just to be sure, I do the "Synthesize | Check syntax" and that one complets with the expected green check mark.
I'm kinda stumped. What am I missing here?
Solved! Go to Solution.
03-03-2009 10:11 PM
Gee - don't everybody reply at once ...
Here's more details: on my work-box (a P4 2.8, running Win2k3) I can install the webPack 10.1 and I can synthesize my design just fine. No errors, no warnings, synthesis gives me a green checkmark. Unfortunately that box doesn't have a parallel port for my dinky JTAG cable.
At home I have a P4-3.6, running WinXPPro (SP3). I installed the same WebPack 10.1. The same project with the same source files fails synthesis. No errors. No warnings. Just "Process Synthesis Fails" as the last line output. I uninstall the whole thing, re-install, same behaviour.
On neither machine did I set any particular flags or anything.
Given that this seems to be the "default behaviour", I don't see how I can be the only person who's ever encountered this. Is there really nobody out there with a hint or something?
03-04-2009 06:15 AM
03-05-2009 11:56 AM
Heh -- thanks for at least thinking about it.
I get the same result for a File|New project, File|Open Project and even File|Open Example. Things seem to synthesize fine and then the last line is "Process synthesis failed" and in the Processes window I get the red "x".
When I copy the output down in the console, I find that it is character-by-character identical on both machines with the exception of the last line (which is either "failed" or "succeeded" of course).
The summary html files are identical except for the time/date stamps.
In particular I'd like to note that the summary html file claims "Current state: synthesized" -- so whatever fails happens after that html file is generated.
Closing a fully synthesized/implemented project with all green checkmarks and re-opening opens it into the same status of "all green checkmarks" - but if I copy the whole directory onto a USB stick and transfer it to my home box and open it there is opens in the "orange question mark" state.
Last night the "Xilinx update" had some new things to offer so I let it update all its stuff but it didn't make any difference in the observed behaviour.
(Just to be complete in my error reporting).
03-06-2009 10:24 PM
I got it.
I figured I'd have a little look at the code itself -- how hard can it be. Boy, the ISE is pretty amateurish hackery -- and trust me, I'm already being pretty polite when I say that.
Anyways: in the file
there is a function dpm_parseOutputFile (a little past line 900) in which the programmers cleverly commit this abomination:
set sFile [join $lFile " "]
which will strip out any redundant whitespace from file- or directory names, of course. Thus if someone has a directory whose name consists of more than one word and which has more than one space separating the words (as in "two spaces", for example) the thusly constructed filename will only have one space at that place and thus the files in that directory will not be found.
You know what this will do to multi-byte characters in directory names? Like, say, directory names in Japanese? Unicode is how old now? and TCL is fully unicode aware! And the ISE programmers insist on breaking it!
Really -- this is the kind of code I wrote back in the middle nineties. Xilinx should shoot its developers and hire someone who knows something about computers...