Christian Maeder wrote:

...

RtsUtils.c:190: error: too few arguments to function 'ctime_r'

I could carry on after adding an argument ", 26"

now I get an error when linking the stage2 compiler. How should I fix this? Cheers Christian /home/maeder/haskell/solaris/ghc-6.4.2.20060411/ghc/rts/libHSrts_thr.a(OSThreads .thr_o): In function `yieldThread': OSThreads.c:(.text+0x88): undefined reference to `sched_yield' collect2: ld returned 1 exit status <> gmake[2]: *** [stage2/ghc-6.4.2.20060411] Error 1 gmake[2]: Leaving directory `/home/maeder/haskell/solaris/ghc-6.4.2.20060411/ghc /compiler' gmake[1]: *** [stage2] Error 2 gmake[1]: Leaving directory `/home/maeder/haskell/solaris/ghc-6.4.2.20060411' gmake: *** [bootstrap2] Error 2 (

Simon Marlow wrote:

GhcBinDistDirs is set by ghc/mk/config.mk, which is included by the top-level Makefile.

I've no such variable in ghc/mk/config.mk or ghc/mk/config.mk.in C.

Christian Maeder wrote:

Simon Marlow wrote:

...

GhcBinDistDirs is set by ghc/mk/config.mk, which is included by the top-level Makefile.

I see, there's another mk/config.mk in the subdirectory "ghc"

...

$ make show Project=Ghc VALUE=GhcBinDistDirs GhcBinDistDirs="ghc libraries hslibs"

in this subdirectory I get the same result, but 'GhcBinDistDirs=""' one level up.

This bit of Makefile code is possibly going wrong for you: -include $(shell echo $(ProjectDirectory) | tr A-Z a-z)/mk/config.mk -include $(shell echo $(ProjectDirectory) | tr A-Z a-z)/mk/version.mk Try this instead: $ make Project=Ghc ProjectDirectory=ghc show VALUE=GhcBinDistDirs from the top level. Maybe that tr command doesn't work on Solaris? Cheers, Simon

Volker Stolz wrote:

venus [14:45:44]> echo foo | tr a-z A-Z foo venus [14:46:11]> echo foo | tr [a-z] [A-Z] FOO [..] Maybe it's safe to switch to the bracketed version these days... This also works fine on Linnix and FreeBSD. Finding the correct 'tr' on Solaris is probably more tedious.

the bracketed version works for me, too. But maybe such a range is wrong for different locales anyway? ([A-Z] in file patterns on my linux machine also match lower case letters except "a"). I'm quite confused by the various ways of doing such a trivial transformation (using perl, sed, tr and what not.) and then still having it wrong for certain machines or environments. Cheers Christian

Christian Maeder wrote:

Simon Marlow wrote:

...

This bit of Makefile code is possibly going wrong for you:

-include $(shell echo $(ProjectDirectory) | tr A-Z a-z)/mk/config.mk -include $(shell echo $(ProjectDirectory) | tr A-Z a-z)/mk/version.mk

you're right!

...

Try this instead:

$ make Project=Ghc ProjectDirectory=ghc show VALUE=GhcBinDistDirs

from the top level.

This shows the values correctly now.

...

Maybe that tr command doesn't work on Solaris?

When I replace "tr A-Z a-z" with

tr "[:upper:]" "[:lower:]"

(taken from "man tr') 'gmake binary-dist' works for me now, too.

I'm rolling and uploading the 6.4.2 bits now, and this patch didn't make it in, sorry. We can put together a patchset for building on Solaris and make it available separately. Can you send me a working patch? Did the DriverState patch I posted earlier work? Cheers, Simon

Christian Maeder wrote:

Simon Marlow wrote:

...

I'm rolling and uploading the 6.4.2 bits now, and this patch didn't make it in, sorry. We can put together a patchset for building on Solaris and make it available separately. Can you send me a working patch? Did the DriverState patch I posted earlier work?

I've integrated no patch from you. Which post was it.

This one: http://www.haskell.org//pipermail/glasgow-haskell-users/2006-April/010002.ht...

A further problem occured when installing my binary distribution. The installed file "bin/ghc" (and bin/ghci) did not have the mini-driver, but only the settings up to DEFAULT_TMPDIR.

Did you have BIN_DIST=1 in your build.mk from the beginning?

I'll rebuild ghc-6-4.2 if you put it out.

Ok, thanks. Simon

Christian Maeder wrote:

The files in the binary distribution are:

ghc -> ghc-6.4.2 ghc-6.4.2

due to a missing dvips "make binary-dist" does not finish properly and does not create a ghc-6.4.2.sh (and other .sh) file(s). Should it not be possible to build a binary distribution without documentation? Cheers Christian hslibs.dvi -o hslibs.ps gmake[2]: hslibs.dvi: Command not found gmake[2]: *** [hslibs.ps] Error 127 gmake[1]: *** [install-docs] Error 1 gmake[1]: Leaving directory `/local/home/maeder/ghc-6.4.2/hslibs' gmake: *** [binary-dist] Error 2

Simon Marlow wrote:

Christian Maeder wrote:

...

Should it not be possible to build a binary distribution without documentation?

Try

make binary-dist Project=Ghc BINDIST_DOC_WAYS=html

Thanks, this worked! C.

* Christian Maeder :

I also had to adapt again the call of ctime_r in ghc/rts/RtsUtils.c

#if HAVE_CTIME_R - ctime_r(&now, nowstr, 26); + ctime_r(&now, nowstr); #else

26 is supposed to be the size of the buffer nowstr. Maybe 27 is also fine? According to the man pages ctime_r is different under linux and solaris. How should that be patched?

Without the 3rd arg is according to POSIX (check your Solaris man page). 'buf' is required to be at least 26 bytes. -- http://www-i2.informatik.rwth-aachen.de/stolz/ *** PGP *** S/MIME "All the excitement lies in pattern matching." (SPJ et al.)

Am 20. Apr 2006 um 11:19 CEST schrieb Christian Maeder:

Volker Stolz wrote:

...

Without the 3rd arg is according to POSIX (check your Solaris man page). 'buf' is required to be at least 26 bytes.

You're right, but it does not compile without the 3rd argument! Maybe this is due to gcc_4.0.3_s10 ? Should one simply rely on ctime?

From my understanding of the patch above (*removing* the 3rd arg), I assumed

Now I'm confused...or is the patch reverted? It said: #if HAVE_CTIME_R - ctime_r(&now, nowstr, 26); + ctime_r(&now, nowstr); #else This has got nothing to do with GCC, it's only related to system headers. Quoth SunOS 5.9: SYNOPSIS #include char *ctime_r(const time_t *clock, char *buf, int buflen); ... POSIX cc [ flag... ] file... -D_POSIX_PTHREAD_SEMANTICS [ library... ] char *ctime_r(const time_t *clock, char *buf); that -D_POSIX_PTHREAD_SEMANTICS was set. It should probably be set like somewhere else (in libraries/bases/cbits/dirUtils.c), but I strongly suggest not doing this in the same (ie. through #define), but rather indeed setting it as a compiler flag. Otherwise, one of those days we'll get bitten by compiling something with skewed headers. Due to the plethora of variables in rts/Makefile, I'm a bit at a loss where to stick this in. -- Volker Stolz * http://www-i2.informatik.rwth-aachen.de/stolz/ * PGP * S/MIME

* Christian Maeder :

P.S. the non-termination of the final stage2 compiler still needs further investigation (that I cannot do alone)

How about throwing 'truss' at it? Maybe this reveals a blatant reason. Volker -- http://www-i2.informatik.rwth-aachen.de/stolz/ *** PGP *** S/MIME FIFA go home!

On 24 April 2006 15:56, Christian Maeder wrote:

Simon Marlow wrote:

...

...

...

Ok, does this help instead (compile stage1 with this change):

*** DriverState.hs.~1.116.2.2.~ 2005-10-13 10:02:19.000000000 +0100 --- DriverState.hs 2006-04-13 15:32:02.000000000 +0100 *************** *** 418,423 **** --- 418,425 ---- #if defined(freebsd_TARGET_OS) "-optc-pthread" , "-optl-pthread" + #elif defined(solaris_TARGET_OS) + "-optl-lrt" #endif ] ),

This patch did not help. No optl option is added anywhere. Maybe "WayThreaded" is the wrong position?

oops, it should be "solaris2_TARGET_OS", not "solaris_TARGET_OS"

This patch works now, with "solaris2" (although I did not see where "-lrt" was finally added.)

Christian

P.S. the non-termination of the final stage2 compiler still needs further investigation (that I cannot do alone)

The best way to proceed would be to run the testsuite with the stage 1 compiler. Grab the test suite from here: http://www.haskell.org/ghc/dist/6.4.2/ghc-testsuite-6.4.2.tar.gz unpack it into your 6.4.2 build tree, cd testsuite, make boot, cd tests/ghc-regress, make 2>&1 | tee log. Cheers, Simon