./configure IGNORES --with-gcc=gcc-4.8, can't build head on os x 10.8 with xcode 5 installed

Im seeing the build scripts using gcc instead of gcc-4.8

this is problematic! with xcode 5, gcc IS an old version of clang!

attaching a copy of the build log.

i'm using 7.6.3 to build this, and the ghc 7.6.3 settings file is pointing to gcc-4.8, so it can only be the build system. I think

[("GCC extra via C opts", " -fwrapv"),
 ("C compiler command", "gcc-4.8"),
 ("C compiler flags", " -m64 -fno-stack-protector  -m64"),
 ("ar command", "/usr/bin/ar"),
 ("ar flags", "clqs"),
 ("ar supports at file", "@ArSupportsAtFile@"),
 ("touch command", "touch"),
 ("dllwrap command", "/bin/false"),
 ("windres command", "/bin/false"),
 ("perl command", "/usr/bin/perl"),
 ("target os", "OSDarwin"),
 ("target arch", "ArchX86_64"),
 ("target word size", "8"),
 ("target has GNU nonexec stack", "False"),
 ("target has .ident directive", "True"),
 ("target has subsections via symbols", "True"),
 ("LLVM llc command", "llc"),
 ("LLVM opt command", "opt")
 ]

is host 7.6.3 ghc setup

Trac metadata

Trac field	Value
Version	7.7
Type	Bug
TypeOfFailure	OtherFailure
Priority	highest
Resolution	Unresolved
Component	Compiler
Test case
Differential revisions
BlockedBy
Related
Blocking
CC
Operating system
Architecture

mentioned in issue #7684 (closed)

mentioned in issue #8143 (closed)

mentioned in issue #8145 (closed)

changed milestone to %7.8.1

changed weight to 10

added Tbug Trac import labels

changed title from ./configure IGNORES --with-gcc=gcc-4.8 to ./configure IGNORES --with-gcc=gcc-4.8, can't build head on os x 10.8 with xcode 5 installed

Attached file run.txt (download).

run of build

if i make a symlink from gcc to whatever gcc i actually want, the problem goes away.

seems that theres some cruft in the build system!

namely, some build scripts for building stage 1 ignore the --with-gcc=somename ignores somename, and calls gcc!

see ticket #8143 (closed) which led to me creating this ticket

i'm then still having some cpp issues.

lots of errors of the form

   Not in scope: ‛unI’
    Perhaps you meant ‛unI#’ (line 111)

the relevant code is

#define derivePrim(ty, ctr, sz, align, idx_arr, rd_arr, wr_arr, idx_addr, rd_addr, wr_addr) \
instance Prim ty where {                                        \
  sizeOf# _ = unI# sz                                           \
; alignment# _ = unI# align                                     \
; indexByteArray# arr# i# = ctr (idx_arr arr# i#)               \
; readByteArray#  arr# i# s# = case rd_arr arr# i# s# of        \
                        { (# s1#, x# #) -> (# s1#, ctr x# #) }  \
; writeByteArray# arr# i# (ctr x#) s# = wr_arr arr# i# x# s#    \
                                                                \
; indexOffAddr# addr# i# = ctr (idx_addr addr# i#)              \
; readOffAddr#  addr# i# s# = case rd_addr addr# i# s# of       \
                        { (# s1#, x# #) -> (# s1#, ctr x# #) }  \
; writeOffAddr# addr# i# (ctr x#) s# = wr_addr addr# i# x# s#   }

unI# :: Int -> Int#
unI# (I# n#) = n#

does that mean the stringification patches aren't in the apple clang? Does this cpp macro work correctly with clang HEAD?

this is in the Types.hs file of the Primitive(s) library

when i do ./configure --with-gcc=gcc-4.8 plus the gcc symlink the build succeeds past this point.

i'll try with clang head next for that related OS X need

Thanks. The GCC invocation is certainly a bug; we definitely want to use the configured C compiler.

As for Clang, can you please test this:

$ clang -E -undef -traditional -o - -
#define FOOBAR(a) test(a, # a)
FOOBAR(xyz)
^D

Please specify the exact path to your clang to be 100% sure. This tests the behavior of hashes in -traditional mode.

If everything is OK, you should see output like this:

# 1 "<stdin>"
# 1 "<command-line>"
# 1 "<stdin>"

test(xyz, # xyz)

If you see output where the hash operator stringifies the argument, then there is a problem.

carter Desktop/repoScratcher » clang -E -undef -traditional -o - -
 #define FOOBAR(a) test(a, # a)
 FOOBAR(xyz)
 ^D
^K^C
carter Desktop/repoScratcher » clang -E -undef -traditional -o - -                                                           130 ↵
#define FOOBAR(a) test(a, # a)
FOOBAR(xyz)
# 1 "<stdin>"
# 1 "<built-in>" 1
# 1 "<built-in>" 3


# 1 "<command line>" 1
# 1 "<built-in>" 2
# 1 "<stdin>" 2

test(xyz, "xyz")

is that the "right output" ?

carter Desktop/repoScratcher » clang --version
Apple LLVM version 5.0 (clang-500.1.63) (based on LLVM 3.3svn)
Target: x86_64-apple-darwin12.4.0
Thread model: posix 

is the clang version i have, which is the newest xcode 5 dev preview. and an apple engineer has explicitly said that is HAS BOTH your clang HEAD patches.

i'll try out the full path stuff tomorrow

i tried with the full path, it has the same problem wrt primitive unI# operation CPP

I'm trying

 CC=pathtoclang ; ./configure --with-gcc=pathtoclang  

right now.

it fails also

I'm going to install the newest xcode cli tools and try that, and if that fails, build clang head and try that

ok, i managed to build stage2 successfully with apple clang from apple dev tools xcode 5 preview, the newest version made available as of august 21st. (ie the version released yesterday).

I used

CC=pathtoclang ; ./configure --with-gcc=pathtoclang

(note: i should have treally tried it without the CC=pathtoclang, but i've not had the time )

I'm running the test suite now. I'm seeing some funny errors, I'll attache the reports in a little while once the test suite ends. (its taking a while to run it all!)