Re: [PATCH] Add Tru64/OSF1 support in Makefile
- From
- Brandon Casey <brandon.casey.ctr@nrlssc.navy.mil>
- Date
- Mar 31, 2010, 16:29 UTC
- Message-ID
- <DA8Or5TA0_betHA0BMcLIT0so7oVjDK6CYr4x1LCiX60SEmuPqVnXg@cipher.nrlssc.navy.mil>
- In-Reply-To
- <d2d39d861003301122p646efc1fj9d88af1893b84b3c@mail.gmail.com>
On 03/30/2010 01:22 PM, Tor Arntsen wrote:
Show 49 quoted lines
> This patch to Makefile makes git build for me on Tru64/OSF1 V5.1 with > the native compiler. > > It needs libcurl installed either in /usr or /usr/local, and ditto for > zlib. The zlib installed on my old Tru64 box is from the original > 'apt-like' tachometer repository at hp.com, now gone. That version is > pretty old, so my patch includes 'NO_DEFLATE_BOUND'. tcl/tk (included > standard on Tru64) is 8.2 which is too old, so that one (and thus > 'gitk') is also disabled in the patch. If someone has V5.2 or newer > they could check tcl/tk version etc., add some $(uname_R) directives > and adjust as necessary. > > The resulting git version is fully working (as far as I can tell), and > now in use for the Tru64 autobuilds at http://curl.haxx.se/auto/ > (which just does 'git pull' regularly). And the patch is produced > locally on the Tru64 machine, with git, obviously. > > Patch included below, as well as attached (in case of > mailler-mangling). Please let me know if you feel there are any issues > with it. > > -Tor > > From d4a5ba0727d7e3a4455320bad641cb34402d16fd Mon Sep 17 00:00:00 2001 > From: Tor Arntsen <tor@spacetec.no> > Date: Tue, 30 Mar 2010 19:36:40 +0200 > Subject: [PATCH] Add Tru64/OSF1 support in Makefile > > Tested with V5.1 > Needs libcurl and zlib in /usr/local. zlib may be e.g. the original > Tru64 tachometer-freeware version (1.1.4) or newer. > > Signed-off-by: Tor Arntsen <tor@spacetec.no> > --- > Makefile | 26 ++++++++++++++++++++++++++ > 1 files changed, 26 insertions(+), 0 deletions(-) > > diff --git a/Makefile b/Makefile > index 3a6c6ea..8041bde 100644 > --- a/Makefile > +++ b/Makefile > @@ -758,6 +758,32 @@ ifeq ($(uname_S),UnixWare) > NO_STRCASESTR = YesPlease > NO_MEMMEM = YesPlease > endif > +ifeq ($(uname_S),OSF1) > + # Tested with V5.1 w/libcurl and zlib-1.1.4-5 in /usr/local/ > + CC = cc > + CFLAGS = -O2 -g3
I wonder if the native compiler will be the most common compiler used on this platform? I tend to lean towards configuring the Makefile for gcc, unless gcc is not available for a platform. The user can set CC and CFLAGS in config.mak and on the make command line if they desire something different.
This isn't an objection by me, I'm just mentioning it to bring attention to the issue in case others think it is worthy for git to have a stated policy one way or the other.
> + SHELL_PATH = /bin/ksh
Can you run the test suite with this ksh? If it suffers from the same exit status behavior of ksh on IRIX and SunOS, then I have a patch which allows the test suite to run.
Also, I wonder if it suffers from the same trap handling that ksh has on SunOS.
> + BASIC_CFLAGS += -D_POSIX_C_SOURCE=1 -D_OSF_SOURCE=1 > + BASIC_CFLAGS += -pthread
There is a PTHREAD_LIBS variable and it seems there is precedent for setting it to -pthread in the FreeBSD section.
Namely, the following shell snippet should exit 0, but does not on SunOS's ksh:
(atrap () { exit 0; }
func3 () { exit 1; }
func2 () { func3; }
func1 () { trap atrap EXIT; func2; }
func1)-brandon
Show 22 quoted lines
> + BASIC_CFLAGS += -I/usr/local/include > + BASIC_LDFLAGS += -L/usr/local/lib > + NO_STRCASESTR = YesPlease > + NO_STRTOUMAX = YesPlease > + NO_STRTOULL = YesPlease > + NO_STRLCPY = YesPlease > + NO_MEMMEM = YesPlease > + NO_PYTHON = YesPlease > + NO_NSEC = YesPlease > + NO_OPENSSL = YesPlease > + NO_UINTMAX_T = YesPlease > + NO_C99_FORMAT = YesPlease > + NEEDS_LIBICONV = YesPlease > + NO_DEFLATE_BOUND = YesPlease > + # V5.1 has tcl/tk 8.2, which doesn't work. > + # V5.2 could possibly have a working version (8.4 or newer). > + NO_TCLTK = YesPlease > + NO_EXPAT = YesPlease > +endif > ifeq ($(uname_S),SCO_SV) > ifeq ($(uname_R),3.2) > CFLAGS = -O2