tech-toolchain archive

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]

HEAD gcc build failures, lex vs. flex



Hi,
I recently switched my nightly builds over to a Linux cluster I had lying around and the builds broke. It looked like GCC was trying to build with flex and it wasn’t trying to use nblex in the tools collection. 

I dug into it a bit and I saw the following in binutils/config.log 

  configure:7866: checking for flex
  configure:7893: result: /vol/build/cylc/20181013T2300+13/build-HEAD/tools.i386/bin/nblex
  configure:7914: checking for flex
  configure:7944: result: no

Well, that’s odd. Checking for flex twice and only finding it once.

It turns out that the first check is for flex/lex, it allows the user to force a program by setting the shell variable LEX and it sets the result variable ac_cv_prog_LEX. 

The second check seems to be for flex alone, it allows the user to force a program by setting the shell variable FLEX and it sets the result variable ac_cv_prog_FLEX. 

I bet we are setting LEX, but GCC really wants FLEX to be set. 

In the file tools/Makefile.gnuhost, I went and added the line "FLEX=${LEX:Q} \” after the line that says "LEX=${LEX:Q} \” and my build seems to be working much better now.

Cheers,
Lloyd


Home | Main Index | Thread Index | Old Index