Compiler flags for GCCFSS to support POSIX & Solaris threads

My team is working on compiling a custom Perl with support for both POSIX and Solaris threads, and I'm trying to work out the appropriate compiler and linker flags to set for GCCFSS. At present I've come up with:
gcc -mt -lpthread -D_POSIX_C_SOURCE=199506L
which expands to:
gcc -D_REENTRANT -lthread -lpthread -D_POSIX_C_SOURCE=199506L
According to the [Solaris 9 Multithreaded Programming Guide|http://docs.sun.com/app/docs/doc/806-6867/6jfpgdcob?a=view], this should allow for both POSIX and Solaris threads support via POSIX semantics.
Any comments or recommendations on this?
Thanks,
RT.

It sounds about right. The interoperability question may be better posted on the Solaris forum though.
Thanks
Prashanth

Similar Messages

Compile flag for the endian neutral conversion on x86_x64 systems

I am struggling to work on the endianess issue porting the driver and application codes from Sparc to x64 system. The device driver just copies the data in and out the structure from the kernel space to user space or from the user space to kernel space calling ddi_copyin and ddi_copyout in most time. The application ported from Sparc has to be handle the endianess issue.
For instance, the driver returns the data in the structure
typedef struct {
ushort robot_address; /* medium transport element address */
ushort robot_count; /* number medium transport elements */
ushort cell_address; /* medium storage element address */
ushort cell_count; /* number medium storage elements */
ushort port_address; /* import-export element address */
ushort port_count; /* number import-export elements */
ushort drive_address; /* data transfer element address */
ushort drive_count; /* number data transfer elements */
} element_info_t;
Byte _0 _1 _2 _3 _4 _5 _6 _7 _8 _9 A B C D E F
00 01 00 01 10 00 00 17 00 10 00 01 01 00 00 02
First Robot Address ........... 256
Number of Robots .............. 256
First Cell Address ............ 16
Number of Storage Cells ....... 5888
First Import/Export Address ... 4096
Number of Import/Exports ...... 256
First Drive Address ........... 1
Number of Drives .............. 512
The correct value for First Robot Address should be 1 in Byte[0]->Byte[1], but the value is read to 256 in Byte[1]->Byte[0] since of the little endian. To get the correct value, I have to create a micro to convert the data, but it will take a lot of time to add the micro in the codes with 20k lines.
Is any compile flag applied to convert the little endian to big endian to save my life for the porting?
I am using Sun Studio 11 with Sun C 5.8 Patch 121016-02 2006/03/31.
Thanks for your help!

There is no compiler option to control endianness of programs. I don't offhand see how such an option could work, because some data will always be in the machine's native endianness.
But there is a standard Unix toolset "xdr" that allows you to send data across an interface between different architectures, while preserving the data values. Each system knows its native data format. The sender marshalls the data stream into the interface format, and the receiver unmarshalls the stream into its own format.
("Every problem in computer science can be solved by adding a level of indirection.")
Run "man xdr" as a place to start.

Perl 5.10 64bit compile fails with gccfss 4.2 on Solaris 10 (5/08) SPARC

Hi ,
I am having problems building my own perl 5.10 using the new GCCFSS 4.2 compiler running Soalris 10(5/08) on SUN T5120 server. I get pass configure but fails at make, below are my configure options and the make error msg. Any help would be appreciated and I have been working on this problem for hours.
Thanks,
Kaprino
sh Configure -Dcc="gcc -m64 -B/usr/ccs/bin/" -Dmake=/usr/css/bin/make -Dusethreads -Uuselargefiles -Duse64bitall -Dusesocks \
-Dprefix=/opt/perl-5.10.0 \
-Aldflags="-mcpu=v9 -m64" \
-Alddlflags="-mcpu=v9 -m64 -G" -des
#### Make Error message #####
SOCKSinit perl.o
ld: fatal: Symbol referencing errors. No output written to miniperl
collect2: ld returned 1 exit status
*** Error code 1
make: Fatal error: Command failed for target `miniperl'
Latest Compiler version
gcc -vUsing built-in specs.
Target: sparc-sun-solaris2.10
Configured with: /net/clpt-v490-1/export/data/bldmstr/20080102_phobos_gcc42/src/configure prefix=/usr/sfw enable-shared with-system-zlib enable-checking=release disable-libmudflap enable-languages=c,c++ enable-version-specific-runtime-libs with-cpu=v9 with-ld=/usr/ccs/bin/ld without-gnu-ld disable-gnattools with-mpfr=/ws/gccfss/tools --with-gmp=/ws/gccfss/tools
Thread model: posix
gcc version 4.2.0 (gccfss)

Hi All,
Thanks for ur reply, but still im getting the below error
*** Error code 1
The following command caused the error:
test ! -f gcc/Makefile || exit 0; \
[ -d gcc ] || mkdir gcc; \
r=`${PWDCMD-pwd}`; export r; \
s=`cd .; ${PWDCMD-pwd}`; export s; \
CC="gcc"; export CC; \
CFLAGS="-g -O2"; export CFLAGS; \
CONFIG_SHELL="/bin/sh"; export CONFIG_SHELL; \
CXX="c++"; export CXX; \
CXXFLAGS="-g -O2"; export CXXFLAGS; \
TOPLEVEL_CONFIGURE_ARGUMENTS="configure --prefix=/usr/sfw --enable-shared --with-system-zlib --enable-checking=LIST --disable-libmudflap -enable-language=c,c++--enable-version-specific-runtime-libs --with-cpu=v9-m64-Aldflags=-mcpu=v9-m64-Alddlflags=-mcpu=v9-m64-G --with-ld=/usr/ccs/bin/ld --without-gnu-ld --disable-gnattools --with-mpfr=/ws/gccfss/tools --with-gmp=/ws/gccfss/tools"; export TOPLEVEL_CONFIGURE_ARGUMENTS; \
AR="ar"; export AR; \
AS="as"; export AS; \
CC_FOR_BUILD="gcc"; export CC_FOR_BUILD; \
DLLTOOL="dlltool"; export DLLTOOL; \
LD="/usr/ccs/bin/ld"; export LD; \
NM="nm"; export NM; \
RANLIB="ranlib"; export RANLIB; \
WINDRES="windres"; export WINDRES; \
OBJCOPY=""; export OBJCOPY; \
OBJDUMP=""; export OBJDUMP; \
SET_GCC_LIB_PATH_CMD="LD_LIBRARY_PATH=/dumps/gcc/gcc-3.4.6/gcc; export LD_LIBRARY_PATH;"; export SET_GCC_LIB_PATH_CMD; \
LD_LIBRARY_PATH=/dumps/gcc/gcc-3.4.6/gcc; export LD_LIBRARY_PATH; \
echo Configuring in gcc; \
cd gcc || exit 1; \
case . in \
srcdiroption="--srcdir=."; \
libsrcdir=".";; \
/* | [A-Za-z]:[\\/]*) \
srcdiroption="--srcdir=./gcc"; \
libsrcdir="$s/gcc";; \
srcdiroption="--srcdir=.././gcc"; \
libsrcdir="$s/gcc";; \
esac; \
/bin/sh ${libsrcdir}/configure \
cache-file=./config.cache build=sparc-sun-solaris2.10 host=sparc-sun-solaris2.10 target=sparc-sun-solaris2.10 prefix=/usr/sfw enable-shared with-system-zlib enable-checking=LIST disable-libmudflap -enable-language=c,c++enable-version-specific-runtime-libs with-cpu=v9-m64-Aldflags=-mcpu=v9-m64-Alddlflags=-mcpu=v9-m64-G with-ld=/usr/ccs/bin/ld without-gnu-ld disable-gnattools with-mpfr=/ws/gccfss/tools with-gmp=/ws/gccfss/tools with-gcc-version-trigger=/dumps/gcc/gcc-3.4.6/gcc/version.c program-transform-name='s,y,y,' ${srcdiroption} \
|| exit 1
make: Fatal error: Command failed for target `configure-gcc'
pls help me.
Regards,
Suresh.
Edited by: ssubrama on Mar 4, 2009 1:18 AM

Compiling mod_perl for Solaris 10 .... grrrrrr

I've got a scenario where I need to run MySQL in a cluster, and the application requires curl be compiled into php. After chasing my tail for weeks trying to get things to play nicely together, I decided to just compile everything from scratch. I have several identical boxes (V125) that need the same software stack also, so rather than building each one manually, I'm attempting to build things as statically as possible so that I can just tar up the stack and distribute it to all the machines that way.
apache 2.0.59
php 5.2.0
perl 5.8.8
MySQL 5.0.27
openssl 0.9.8d
curl 7.15.5
expat 2.0.0
gettext 0.16.1
libxml2 2.6.27
libiconv 1.11
readline 5.2
zlib 1.2.3
And several perl modules, including DBI/DBD for both Oracle and MySQL
I used the gcc compiler and associated environment packages from Blastwave. In the end I had a virtually static setup (still relies on things from /lib which is fine by me), except for the php5 and mod_perl modules for apache. The php worked like a champ as a loadable, but apache dies when trying to load the perl module:
Syntax error on line 233 of /opt/CA/apache2.0.59/conf/httpd.conf:
Cannot load /opt/CA/apache2.0.59/modules/mod_perl.so into server: ld.so.1: httpd: fatal: relocation error: file /opt/CA/apache2.0.59/modules/mod_perl.so: symbol PL_markstack_ptr: referenced symbol not found
I really don't know how to troubleshoot this any further. Any ideas or help would be appreciated. I will attach my compile strings below so you can see what I have done.
./configure prefix=/opt/CA/gettext-0.16.1 program-transform-name='s/^gettext$/g&/;s/^msgfmt$/g&/;s/^xgettext$/g&/' disable-shared enable-static
libiconv
./configure prefix=/opt/CA/libiconv-1.11 disable-shared enable-static with-gettext=/opt/CA/gettext
xml2
LD_OPTIONS="-L/opt/CA/gettext -L/opt/CA/libiconv -L/opt/CA/libiconv/lib -L/opt/CA/gettext/lib"
./configure prefix=/opt/CA/libxml2-2.6.27 disable-shared --enable-static
readline
/configure prefix=/opt/CA/readline-5.2 disable-shared --enable-static
Zlib (could not find any static compile options for this)
./configure prefix=/opt/CA/zlib-1.2.3 shared
openssl
./config prefix=/opt/CA/openssl-0.9.8d openssldir=/opt/CA/openssl-0.9.8d no-shared
apache
./configure prefix=/opt/CA/apache2.0.59 enable-most with-ssl=/opt/CA/openssl disable-shared --enable-static
CFLAGS="-mimpure-text"
./configure prefix=/opt/CA/php5.2.0 enable-bcmath enable-wddx with-mysql=/opt/CA/mysql with-apxs2=/opt/CA/apache/bin/apxs with-apache2=/opt/CA/apache enable-force-cgi-redirect enable-discard-path with-config-file-path=/opt/CA/php-5.2.0/etc enable-safe-mode enable-libgcc with-openssl=/opt/CA/openssl with-zlib=/opt/CA/zlib with-zlib-dir=/opt/CA/zlib with-readline=/opt/CA/readline without-pear with-curl=/opt/CA/curl enable-lib-libxml with-iconv=/opt/CA/libiconv with-mysql=/opt/CA/mysql with-mysqli=/opt/CA/mysql/bin/mysql_config with-libxml-dir=/opt/CA/libxml2 enable-shared=no enable-static=yes
mod_perl
/opt/CA/perl/bin/perl Makefile.PL MP_APXS=/opt/CA/apache/bin/apx
Thanks in advance for any and all help!
Dave

Ok, I found my answer. Still don't know how to fix 1.37 itself, but turns out that the problem is fixed in 1.38, for anyone else with this issue.

Getconf returns compiler flags that the compiler warns about

I have Workshop 12u1 installed on my Solaris 10u8 machine (x86_64). I'm up to date with patches as of a couple weeks ago.
I'm well aware that the Workshop suite has deprecated flags like -xarch=generic64, -xarch=amd64a, etc.
However, when I run "getconf" on my Solaris 10 system, some of the conf settings still return those older flags that the compiler now warns about. That causes a lot of warning chatter when building certain applications (e.g. perl 5.10.1) that are smart enough to use "getconf" to try determine what flags to use in certain situations. Using getconf to find the right flags is even mentioned in the standards(5) page, but following its advice results in deprecation warnings from the compiler. Not good.
If you run the following code you'll see what I mean:
for f in `man getconf | egrep 'XBS'`; do echo "Checking $f"; getconf "$f"; doneIs this something that I (as a sysadmin for the system) should be able to configure, or is it something that would need to be patched to fix? The man pages for getconf, sysconf, and confstr aren't clear where the values are actually coming from.
So, how should this be fixed?
Thanks,
Tim

Enchanter wrote:
Since I'm running regular old Solaris 10, rather than OpenSolaris, I'm skeptical how much help the OpenSolaris developers are truly going to be. I can certainly give it a go, but even if they fix it in OpenSolaris, it might be a very long time before that fix filters into Solaris.Actually, it goes the other way. New features and modification go into the internal Solaris development workspace before they migrate to Open Solaris. The Solaris Express and final releases and Open Solaris releases come from the same source base. Whether a feature or update goes into a Solaris 10 patch depends partly on demand, especially demand from customers with service contracts.
I'm also a little puzzled about the "backwards compatibility" thinking of the compiler developers. In some areas (e.g. C++ and the standard library), what's shipped with the compiler is years and years behind what's current, apparently in the name of "backward compatibility". Yet where compiler flags are concerned, the developers seem to be much more cavalier about deprecating things and making changes that break backward compatibility.We try very hard not to break code or makefiles. Sun (and now Oracle) makes its money from enterprise users. Their code bases last for a long time, and changes are expensive. Changing a line in a makefile or source code file can mean having to re-certify the application.
The change from -xarch to -m specifying the memory model was more abrupt than we would have liked, and in hindsight, was probably not handled very well. We were running into combinatorial explosion of -xarch sub-options due to the increasing number of architectures that the compiler supports. Some option combinations quietly resulted in behavior that you didn't expect. (For example, the combination "*-fast -xarch=v9*"did not give the same result as "*-xarch=v9 -fast*".) Separating memory model from the other considerations made the options easier and more reliable to use.
Don't get me wrong -- I'm all for forward progress with the tool-chain -- I think that the C++ standard library and the defaults should all be brought up to what's current in the industry. Sun Studio 12 update 1 has direct support for Apache stdcxx (if installed in a standard location) in addition to libCstd and STLport. The Apache library will be in the next Open Solaris. When it will be available in a Solaris 10 update has not yet been decided.
I also think adding support for -m32/-m64 was a good move. It just seems to me that if the developers are going to deprecate a large number of flags that the compiler accepts, they should also make certain that the rest of the operating system can handle that change. To me, that means that they need to work with the developers of getconf/confstr() to make certain it knows how to check the compiler version before it emits a particular flag.Only a few options have been deprecated. It's just that they are probably the ones you used most. :-(
More than one compiler version can be installed on a system, and compilers can be installed anywhere. Programmers in medium to large shops typically run compilers installed on a remote server, although they might have other compilers installed locally. If getconf had to guess what compiler you were using, it would often guess wrong. If you have suggestions on how to match up getconf behavior with the compiler you want to use, please post them in an Open Solaris forum.

HTTP Server is not currently supported with Solaris 10

I am having a problem to start the HTTP Server on Solaris 10 (I am installing HTTP Server as part of HTMLDB installation on Solaris 10).
I had a problem to install it at first, however I've found on 'AskTom' that unsetting ORA_NLS33 helps and it did. Now I can't start the HTTP server and I've found on metalink that 'HTTP Server is not currently supported with Solaris 10, this is targeted for Q2CY05'.
Has this been fixed yet?
Thank you for your time.
DanielD

Hi james,
I have checked the log file but i am unable to find what to do next.
I am giving the small piece of content of the log file. The following messages are repeating again and again...
Log content:-
07/09/23 17:20:36 Start process
/dev011/oracle/product/10.1.3.1/OracleAS_1/Apache/Apache/bin/apachectl startssl: execing httpd
07/09/23 17:22:26 Start process
/dev011/oracle/product/10.1.3.1/OracleAS_1/Apache/Apache/bin/apachectl startssl: execing httpd
07/09/23 17:22:30 Start process
/dev011/oracle/product/10.1.3.1/OracleAS_1/Apache/Apache/bin/apachectl startssl: execing httpd
07/09/23 17:24:40 Start process
/dev011/oracle/product/10.1.3.1/OracleAS_1/Apache/Apache/bin/apachectl startssl: execing httpd
07/09/23 17:24:44 Start process
/dev011/oracle/product/10.1.3.1/OracleAS_1/Apache/Apache/bin/apachectl startssl: execing httpd
07/09/23 17:26:36 Start process
/dev011/oracle/product/10.1.3.1/OracleAS_1/Apache/Apache/bin/apachectl startssl: execing httpd
07/09/23 17:26:42 Start process
/dev011/oracle/product/10.1.3.1/OracleAS_1/Apache/Apache/bin/apachectl startssl: execing httpd
07/09/23 17:28:57 Start process
/dev011/oracle/product/10.1.3.1/OracleAS_1/Apache/Apache/bin/apachectl startssl: execing httpd
07/09/23 17:29:01 Start process
/dev011/oracle/product/10.1.3.1/OracleAS_1/Apache/Apache/bin/apachectl startssl: execing httpd

Deletion flag for customer

Hi all,
I set a Deletion flag for a customer in xdo6, when i am trying make the sale order system is giving only warnning message and allowing me to make the sale order. But i need a error message and should not allow me to make the sale order , support me to come out this issue.
Thanks in advance

Use VD05 to block the customer for Order, Delivery or billing
Use XD06 for deletion flag.
If your requirement is that orders should not be created for that customer, VD05 will resolve it.
in VD05, put customer number and select Sales area by customer.Here select "Overall Block" for order, delivery and billing.

Essbase Error(1200329): Error compiling formula for [Asset Description] (li

Hi,
I am having some problem in debugging one error
Detail:Cannot calculate. Essbase Error(1200329): Error compiling formula for [Asset Description] (line 37): invalid assignment
I am just creating a Add asset script as my client does not have license to CapEx module.
Script Below- please let me know what can be an issue
VAR numAsset = [NumAsset];
VAR usFulLife=[UsefulLife];
FIX(@CHILDREN("Total New"), [Scenario], [Version], [Entity], [Department], [AssetClass],[Year],"BegBalance","Local", "HSP_InputValue")
"Asset Description" (
IF (numAsset > 0 AND @MAXS(SKIPMISSING, @CHILDREN("Asset Properties")) == #MISSING )
numAsset = numAsset - 1;
          "Asset Status" = 0; /* flag to know this is a new asset*/
          "Asset ID" = [AssetID];
"Asset Description" = [AssetDesc];
          "CAR No." = [AssetCAR];
          "Purchase Units" = [AssetUnits];
          "Purchase Amount(per unit)" = [AssetRate];
          "In Service Date" = [InserviceDate];
     /*     "Total Cost" = "Purchase Units" *"Purchase Amount(per unit)";*/
IF(usFulLife > 0)
          "Useful Life (in Years)" = usFulLife;
ELSE
               "Useful Life (in Years)" = "No Scenario"->"No Version"->"No Entity"->"No Department"->"No LineItem"->"Useful Life (in Years)";
ENDIF
ENDIF
ENDFIX
FIX([Scenario], [Version], [Entity], [Department], "Local", "HSP_InputValue")
CALC DIM ("TimePeriod");
@IANCESTORS("Line Item 1");
@ANCESTORS([AssetClass]);
ENDFIX
Note- "Asset Description" is a member of Text type and in this script [AssetDesc] is a global variable of StringAsNumber
Regards,
-KP

Thanks for responding but my issue resolved, it was a Space in the member name.
-KP

CDE Support on Solaris 8

Hi all !
I have just installed Solaris 8 for Intel on a laptop. After lots of hassel with the video-driver, I got this working by installing Accelerated-X Laptop Display Server. But here it comes .... However,
Xwindows functions OK, whereas CDE fails. Has anyone had similar experience, and more importantly has anyone resolved the problem ?

I am sorry but Workshop 4.2 is supported on Solaris 7 and 8. I don't
see why it shouldn't be. If your statement means that Solaris 7
distribution doesn't come along with Workshop 4.2, then ignore my
statement, I am not sure of that. But Workshop 4.2 will work on both
Solaris 7 and Solaris 8.
Rajag.

JNI FC1063 HBA support on solaris 10

Hi ,
I have upgraded Solaris 9 to Solaris 10 using Live upgrade.Server is SUN fire V880 and the HBA is JNI FC1063.
After upgrading ,I am not able to see the FC LUNS .
Does The JNI Fc1063 HBA has support on Solaris 10.
Thanks

I would beg to differ on this response. This is not a service call and the storage vendor is not responsible for making decisions on what the customer is allowed or not allowed to use.
EG, I have a Sun StorEdge 9990 (actually a HDS USP 1100 and not the 9990 Lightning series which is on the web site) and I made the call to put the Fibre Channel HBA's of my choice into my systems and I also chose the Cisco 9509 switches. Both HDS and Cisco put out documentation regarding compatibility of the HBA's.
So, I would suggest RPS check out the Brocade website or sales people and ask them if they recommend anything.
http://www.brocade.com/products/pdf/Fabric_Aware_Compatibility_Matrix-060521.pdf
states that the option X6768A is supported for Solaris 10.
Sun 9980 will work on most of the HBA's available in the system handbook.
The could be either the rebadged Qlobic or JNI HBA's. You could even use Emulux if you are that way inclined.
My systems are full of dual channel HBA's such as SG-XPCI2FC-QF2.
You should also be aware that the 1280 has only one 66 MHz PCI slot. You can upgrade the I/O boat to all 66 MHz slots if you are after good throughput. The upgrade will also give the opportunity to use the 4 Gb HBA if you can find a system to use them on.
Stephen

Level of ACPI support with Solaris 10?

What is the current level of ACPI support with Solaris 10 x86? I have been searching forums and Sun documentation and I have not found an official answer to this question from Sun. I have seen posts in which people state what they "think" they know or they know how something worked with "their" system but nothing official from Sun. I assume there is some support for ACPI because the eeprom command can be used with the acpi-user-options. Can I just enable ACPI support on my motherboard and let Solaris 10 Update 3 do the rest? Thanks for the help?

I found this site that seams to answer this question.
http://blogs.oracle.com/jsavit/entry/ours_goes_to_11_features

Missing default flag for communication type in crm/business partner

Hello!
We've connected a crm 5.0 with an R/3-system (4.6c).
The crm-system is the leading system...
While inserting a new customer in crm with phone- and fax-number.
After "save" we've got red bdocs, telling us that the standard entry for communication type is missing.
In crm 5.0 there's no standard flag for the phone number but a fixed range with from-date and to-date.
The R/3-system needs the default flag (an "X") as a sign that the number is the standard number.
Anywhere there should be the problem...
Is here anybody with some advices for us?
Thanks a lot in advance.
Best regards,
Ingo

Dear Ingo,
I assume you talk about error message XR002
You have 3 options to resolve the issue:
1. In order to avoid error message XR002 you must set the Standard
Indicator Validity in the maintained telephone in BP transaction.
You can find this indicator in the yellow arrow near the telephone
number. Here you must press the button "Standard Indicator Validity".
2. In customizing you have the possibility to switch off the time
dependency of the addresses causing the XR002.
3. The error occurs due to missing ADRU entries for the communication
data (Email, Phone, Fax etc).
The function modules used in data exchange assumes that there is
an ADRU entry for all addresses. This assumption is partially correct.
If an address is created in Basis release 6.40 and above, then for all
communication data of that address ADRU entries will be created.
But if a Business Partner(with address) is created in release 6.20 or
below and system is upgraded to 7.0, ADRU entries for the communication
data of thoses addresses are not automatically created. When the user
opens that BP in change mode & modifies the communication data and
saves it then ADRU entries for that BP's communication data are created.
There exist a correction report and several notes, so I think if above info
is not able to help, it would be better to open a support message!
Regards,Gerhard

"USE flags" for makepkg

I have an idea for makepkg, but I wanted to discuss it before I file a feature request on flyspry. The idea is similar to Gentoo's USE flags, but don't let this pollude your opinion if you dislike Gentoo.
The idea goes like this. In some config file, the user could specify flags for building, example:
vim=(!perl python ruby !tcl)
This variable would then be exported in the vim PKGBUILD which would interpret it to do:
./configure --disable-perlinterp --enable-pythoninterp --enable-rubyinterp --disable-tclinterp
Makepkg should probably assist in the interpretation somehow.
This is different from Gentoo in the way it is package-specific (allows for better fine tuning) and uses arch-style bash-arrays. Possibly, there could be a global variable aswell (perhaps called ALL or the like).
Pros:
- Easily fine tune packages before building (i.e. yaourt -Sb)
- Don't need to know all the switches for ./configure or the like, to modify a package
Cons:
- Makes creating PKGBUILDs more complex (if the creator wants to support these flags)
- Need to change lots of PKGBUILDS (but this can be done over time)
- ZOMG GENTOO??
There are ofcourse more pros/cons I haven't thought of, and the idea can probably be improved upon - please discuss!

Poisoning the PKGBUILDs with something like this seems like a temporary solution to me. It's nothing that would work in the end and it would generate a big mess. I would not suggest adding a "USE-flag" system to Arch either. It works well in Gentoo, but it was designed with it in mind, Arch was not. It would require some work to implement it.
The best solution I can think of would be to add an optional entry to PKGBUILD, such as optional=('guile' 'tk'), where the optional configure options could be listed. Then makepkg would generate a correct ./configure statement. An example could be like this:
# $Id: PKGBUILD,v 1.8 2007/02/11 11:27:15 damir Exp $
# Maintainer: damir <[email protected]>
# Contributor: damir <[email protected]>
pkgname=xbindkeys
pkgver=1.8.0
pkgrel=1
pkgdesc="xbindkeys allows you to launch shell commands with your keyboard or your mouse under X Window"
arch=(i686 x86_64)
url="http://hocwp.free.fr/xbindkeys"
depends=('libx11')
optional=('guile' 'tk')
source=($url/${pkgname}-${pkgver}.tar.gz)
build() {
cd ${startdir}/src/${pkgname}-${pkgver}
./configure --prefix=/usr ${optional}
make || return 1
make DESTDIR=${startdir}/pkg install
md5sums=('71755c23faa113e891de6f2f871e0c76')
If the user had specified in /etc/makepkg.conf that he/she didn't want tk and guile, makepkg would append --disable-guile --disable-tk to the ./configure statement. If the user want the default, and haven't edited makepkg.conf, nothing will be appended to the ./configure statement.
If the defaults would be not to include guile and tk, then the optional line would look like: optional=('!guile' '!tk')
I didn't give this much thought, so I'm not sure if it would actually work. I also don't know how consistent the application developers name their optional configuration options.

OpenInstaller is only supported on Solaris 9 or later error

I was trying to install java 1.6 version on an Solaris machine During install it failed with the following error message "openInstaller is only supported on Solaris 9 or later" since the Solaris version on this machine is 5.8
As per the sun download page, it was mentioned like supported version is Solaris 8,9 and 10.
http://www.java.com/en/download/help/5000010600.xml
My Solaris machine version is 5.8
./java_ee_sdk-6u1-jdk-solaris-sparc.sh
extracted: install/metadata/view/jwsdp_5.png
extracted: install/metadata/view/jwsdp_6.png
extracted: install/metadata/view/jwsdp_7.png
extracted: install/metadata/view/updatetool.xml
openInstaller is only supported on Solaris 9 or later
Could you please let me know what is the issue and how to resolve this issue.
Thanks,
Ricks

bumping this up in hopes of an answer....
it appears that the jar installer works fine, the problem is with .bin installers only.....
in install docs found proposed solution of "use df compatible with used posix" - that must be a joke, right ??
we're supposed to replace df that comes with OS with some version that works with posix used by the installer ?? Really ?@?@?

Messages are flagged for follow-up, but the user didn't flag them.

Hello All,
I have two users who are complaining that messages they didn't flag are showing up as flagged in their To-Do bars. The odd thing is that the flagged messages are only present if they are not in Cached mode, or if they check OWA. Both users are
also able to access a XenDesktop non-persistent virtual desktop, which presents Outlook in non-Cached mode. When they access Outlook via XenDesktop, the erroneously flagged messages are present. If they use Outlook in Cached mode, the messages
don't appear on their To-Do bar. Our organization doesn't permit PST files, so neither users has a PST. One user mentioned that it he noticed it after he re-ordered some tasks in the To-Do bar by drag-and-drop.
-Outlook 2007
-Exchange 2010 SP3
Multiple Internet searches have found no solutions, so I'm hoping someone in the community here will have an idea.
Thanks!

Hello,
The flag is only an attribute for an email. If online mode and OWA show that the emails are flagged, it means the flag attribute is set to true on the Exchange mailbox.
For this kind of issues, I will check:
1. If the mailbox is shared with other users and the emails are flagged by them. (Any delegates)
2. If there are any third-party add-ins on Outlook that may flagged the emails.
Thanks,
Please remember to mark the replies as answers if they help and unmark them if they provide no help. If you have feedback for TechNet Subscriber Support, contact
[email protected]
Simon Wu
TechNet Community Support

Compiler flags for GCCFSS to support POSIX & Solaris threads

Similar Messages

Maybe you are looking for