In message <15315.4758.978363.964534@iller.crt.se>, you write: 

-> >>>>> matthew green writes:
-> 
-> matthew> true
-> matthew> /usr/src/tools/toolchain/../../gnu/dist/toolchain/ld/ldlex.l
-> matthew> && mv lex.yy.c
-> matthew> /usr/src/tools/toolchain/../../gnu/dist/toolchain/ld/ldlex.c
-> matthew> mv: rename lex.yy.c to
-> matthew> /usr/src/tools/toolchain/../../gnu/dist/toolchain/ld/ldlex.c:
-> matthew> No such file or directory *** Error code 1 [ ... ]
-> 
-> matthew> obviously `true' isn't going to generate lex.yy.c, but why is
-> matthew> this rule being run at all?
-> 
-> I saw this too, and it was caused by "ldlex.l" being newer than
-> "ldlex.[hc]" in .../gnu/dist/toolchain/ld/.
-> 
-> I think using LEX=true and having the already lex'ed file in CVS is
-> probably a good idea, considering cross compiling and stuff, but
-> obviously the "mv lex.yy.c ..." part of the lexer make rule needs to
-> nulled out somehow too.

Just as a "Me TOO!", I've seen this happen on my sgimips boxes on occasion
as well, so it's definitely not just a sparc thing (as Urban's explanation
leads us to believe anyway).

Thanks for doing the digging, Urban... I imagined it was somehow related
to messed up relative times on the files, but never got around to digging
as I was always trying to fix something else at the time.

--rafal

----
Rafal Boni                                                  rafal@mediaone.net