perlhpux.1

视频监控网络部分的协议ddns,的模块的实现代码,请大家大胆指正.

http://www.cmve.net/~merijn/ for HP-UX 10.20, HP-UX 11.00, and HP-UX 11.11
(HP-UX 11i) in both 32\- and 64\-bit versions. These are bzipped tar archives
that also include recent \s-1GNU\s0 binutils and \s-1GNU\s0 gdb.  Read the instructions
on that page to rebuild gcc using itself.
.PP
On PA-RISC you need a different compiler for 32\-bit applications and for
64\-bit applications. On PA-RISC, 32\-bit objects and 64\-bit objects do
not mix. Period. There is no different behaviour for \s-1HP\s0 C\-ANSI-C or \s-1GNU\s0
gcc. So if you require your perl binary to use 64\-bit libraries, like
Oracle\-64bit, you \s-1MUST\s0 build a 64\-bit perl.
.PP
Building a 64\-bit capable gcc on PA-RISC from source is possible only when
you have the \s-1HP\s0 C\-ANSI C compiler or an already working 64\-bit binary of
gcc available. Best performance for perl is achieved with \s-1HP\s0's native
compiler.
.Sh "Using Large Files with Perl on HP-UX"
.IX Subsection "Using Large Files with Perl on HP-UX"
Beginning with HP-UX version 10.20, files larger than 2GB (2^31 bytes)
may be created and manipulated.  Three separate methods of doing this
are available.  Of these methods, the best method for Perl is to compile
using the \-Duselargefiles flag to Configure.  This causes Perl to be
compiled using structures and functions in which these are 64 bits wide,
rather than 32 bits wide.  (Note that this will only work with \s-1HP\s0's \s-1ANSI\s0
C compiler.  If you want to compile Perl using gcc, you will have to get
a version of the compiler that supports 64\-bit operations. See above for
where to find it.)
.PP
There are some drawbacks to this approach.  One is that any extension
which calls any file-manipulating C function will need to be recompiled
(just follow the usual \*(L"perl Makefile.PL; make; make test; make install\*(R"
procedure).
.PP
The list of functions that will need to recompiled is:
  creat,          fgetpos,        fopen,
  freopen,        fsetpos,        fstat,
  fstatvfs,       fstatvfsdev,    ftruncate,
  ftw,            lockf,          lseek,
  lstat,          mmap,           nftw,
  open,           prealloc,       stat,
  statvfs,        statvfsdev,     tmpfile,
  truncate,       getrlimit,      setrlimit
.PP
Another drawback is only valid for Perl versions before 5.6.0.  This
drawback is that the seek and tell functions (both the builtin version
and \s-1POSIX\s0 module version) will not perform correctly.
.PP
It is strongly recommended that you use this flag when you run
Configure.  If you do not do this, but later answer the question about
large files when Configure asks you, you may get a configuration that
cannot be compiled, or that does not function as expected.
.Sh "Threaded Perl on HP-UX"
.IX Subsection "Threaded Perl on HP-UX"
It is possible to compile a version of threaded Perl on any version of
HP-UX before 10.30, but it is strongly suggested that you be running on
HP-UX 11.00 at least.
.PP
To compile Perl with threads, add \-Dusethreads to the arguments of
Configure.  Verify that the \-D_POSIX_C_SOURCE=199506L compiler flag is
automatically added to the list of flags.  Also make sure that \-lpthread
is listed before \-lc in the list of libraries to link Perl with. The
hints provided for HP-UX during Configure will try very hard to get
this right for you.
.PP
HP-UX versions before 10.30 require a separate installation of a \s-1POSIX\s0
threads library package. Two examples are the \s-1HP\s0 \s-1DCE\s0 package, available
on \*(L"HP-UX Hardware Extensions 3.0, Install and Core \s-1OS\s0, Release 10.20,
April 1999 (B3920\-13941)\*(R" or the Freely available \s-1PTH\s0 package, available
on H.Merijn's site (http://mirrors.develooper.com/hpux/).
.PP
If you are going to use the \s-1HP\s0 \s-1DCE\s0 package, the library used for threading
is /usr/lib/libcma.sl, but there have been multiple updates of that
library over time. Perl will build with the first version, but it
will not pass the test suite. Older Oracle versions might be a compelling
reason not to update that library, otherwise please find a newer version
in one of the following patches: \s-1PHSS_19739\s0, \s-1PHSS_20608\s0, or \s-1PHSS_23672\s0
.PP
reformatted output:
.PP
.Vb 10
\&  d3:/usr/lib 106 > what libcma\-*.1
\&  libcma\-00000.1:
\&     HP DCE/9000 1.5               Module: libcma.sl (Export)
\&                                   Date: Apr 29 1996 22:11:24
\&  libcma\-19739.1:
\&     HP DCE/9000 1.5 PHSS_19739\-40 Module: libcma.sl (Export)
\&                                   Date: Sep  4 1999 01:59:07
\&  libcma\-20608.1:
\&     HP DCE/9000 1.5 PHSS_20608    Module: libcma.1 (Export)
\&                                   Date: Dec  8 1999 18:41:23
\&  libcma\-23672.1:
\&     HP DCE/9000 1.5 PHSS_23672    Module: libcma.1 (Export)
\&                                   Date: Apr  9 2001 10:01:06
\&  d3:/usr/lib 107 >
.Ve
.PP
If you choose for the \s-1PTH\s0 package, use swinstall to install pth in
the default location (/opt/pth), and then make symbolic links to the
libraries from /usr/lib
.PP
.Vb 2
\&  # cd /usr/lib
\&  # ln \-s /opt/pth/lib/libpth* .
.Ve
.PP
For building perl to support Oracle, it needs to be linked with libcl
and libpthread. So even if your perl is an unthreaded build, these
libraries might be required. See \*(L"Oracle on HP-UX\*(R" below.
.Sh "64\-bit Perl on HP-UX"
.IX Subsection "64-bit Perl on HP-UX"
Beginning with HP-UX 11.00, programs compiled under HP-UX can take
advantage of the \s-1LP64\s0 programming environment (\s-1LP64\s0 means Longs and
Pointers are 64 bits wide), in which scalar variables will be able
to hold numbers larger than 2^32 with complete precision.  Perl has
proven to be consistent and reliable in 64bit mode since 5.8.1 on
all HP-UX 11.xx.
.PP
As of the date of this document, Perl is fully 64\-bit compliant on
HP-UX 11.00 and up for both cc\- and gcc builds. If you are about to
build a 64\-bit perl with \s-1GNU\s0 gcc, please read the gcc section carefully.
.PP
Should a user have the need for compiling Perl in the \s-1LP64\s0 environment,
use the \-Duse64bitall flag to Configure.  This will force Perl to be
compiled in a pure \s-1LP64\s0 environment (with the +DD64 flag for \s-1HP\s0 C\-ANSI-C,
with no additional options for \s-1GNU\s0 gcc 64\-bit on PA-RISC, and with
\&\-mlp64 for \s-1GNU\s0 gcc on Itanium).
If you want to compile Perl using gcc, you will have to get a version of
the compiler that supports 64\-bit operations.)
.PP
You can also use the \-Duse64bitint flag to Configure.  Although there
are some minor differences between compiling Perl with this flag versus
the \-Duse64bitall flag, they should not be noticeable from a Perl user's
perspective. When configuring \-Duse64bitint using a 64bit gcc on a
pa-risc architecture, \-Duse64bitint is silently promoted to \-Duse64bitall.
.PP
In both cases, it is strongly recommended that you use these flags when
you run Configure.  If you do not use do this, but later answer the
questions about 64\-bit numbers when Configure asks you, you may get a
configuration that cannot be compiled, or that does not function as
expected.
.Sh "Oracle on HP-UX"
.IX Subsection "Oracle on HP-UX"
Using perl to connect to Oracle databases through \s-1DBI\s0 and DBD::Oracle
has caused a lot of people many headaches. Read \s-1README\s0.hpux in the
DBD::Oracle for much more information. The reason to mention it here
is that Oracle requires a perl built with libcl and libpthread, the
latter even when perl is build without threads. Building perl using
all defaults, but still enabling to build DBD::Oracle later on can be
achieved using
.PP
.Vb 1
\&  Configure \-A prepend:libswanted=\*(Aqcl pthread \*(Aq ...
.Ve
.PP
Do not forget the space before the trailing quote.
.PP
Also note that this does not (yet) work with all configurations,
it is known to fail with 64\-bit versions of \s-1GCC\s0.
.Sh "\s-1GDBM\s0 and Threads on HP-UX"
.IX Subsection "GDBM and Threads on HP-UX"
If you attempt to compile Perl with (\s-1POSIX\s0) threads on an 11.X system
and also link in the \s-1GDBM\s0 library, then Perl will immediately core dump
when it starts up.  The only workaround at this point is to relink the
\&\s-1GDBM\s0 library under 11.X, then relink it into Perl.
.PP
the error might show something like:
.PP
Pthread internal error: message: _\|\fI_libc_reinit()\fR failed, file: ../pthreads/pthread.c, line: 1096
Return Pointer is 0xc082bf33
sh: 5345 Quit(coredump)
.PP
and Configure will give up.
.Sh "\s-1NFS\s0 filesystems and \fIutime\fP\|(2) on HP-UX"
.IX Subsection "NFS filesystems and utime on HP-UX"
If you are compiling Perl on a remotely-mounted \s-1NFS\s0 filesystem, the test
io/fs.t may fail on test #18.  This appears to be a bug in HP-UX and no
fix is currently available.
.Sh "perl \-P and // and HP-UX"
.IX Subsection "perl -P and // and HP-UX"
If HP-UX Perl is compiled with flags that will cause problems if the
\&\-P flag of Perl (preprocess Perl code with the C preprocessor before
perl sees it) is used.  The problem is that \f(CW\*(C`//\*(C'\fR, being a \*(C+\-style
until-end-of-line comment, will disappear along with the remainder
of the line.  This means that common Perl constructs like
.PP
.Vb 1
\&  s/foo//;
.Ve
.PP
will turn into illegal code
.PP
.Vb 1
\&  s/foo
.Ve
.PP
The workaround is to use some other quoting separator than \f(CW"/"\fR,
like for example \f(CW"!"\fR:
.PP
.Vb 1
\&  s!foo!!;
.Ve
.Sh "HP-UX Kernel Parameters (maxdsiz) for Compiling Perl"
.IX Subsection "HP-UX Kernel Parameters (maxdsiz) for Compiling Perl"
By default, HP-UX comes configured with a maximum data segment size of
64MB.  This is too small to correctly compile Perl with the maximum
optimization levels.  You can increase the size of the maxdsiz kernel
parameter through the use of \s-1SAM\s0.
.PP
When using the \s-1GUI\s0 version of \s-1SAM\s0, click on the Kernel Configuration
icon, then the Configurable Parameters icon.  Scroll down and select
the maxdsiz line.  From the Actions menu, select the Modify Configurable
Parameter item.  Insert the new formula into the Formula/Value box.
Then follow the instructions to rebuild your kernel and reboot your
system.
.PP
In general, a value of 256MB (or \*(L"256*1024*1024\*(R") is sufficient for
Perl to compile at maximum optimization.
.SH "nss_delete core dump from op/pwent or op/grent"
.IX Header "nss_delete core dump from op/pwent or op/grent"
You may get a bus error core dump from the op/pwent or op/grent
tests. If compiled with \-g you will see a stack trace much like
the following:
.PP
.Vb 10
\&  #0  0xc004216c in  () from /usr/lib/libc.2
\&  #1  0xc00d7550 in _\|_nss_src_state_destr () from /usr/lib/libc.2
\&  #2  0xc00d7768 in _\|_nss_src_state_destr () from /usr/lib/libc.2
\&  #3  0xc00d78a8 in nss_delete () from /usr/lib/libc.2
\&  #4  0xc01126d8 in endpwent () from /usr/lib/libc.2
\&  #5  0xd1950 in Perl_pp_epwent () from ./perl
\&  #6  0x94d3c in Perl_runops_standard () from ./perl
\&  #7  0x23728 in S_run_body () from ./perl
\&  #8  0x23428 in perl_run () from ./perl
\&  #9  0x2005c in main () from ./perl
.Ve
.PP
The key here is the \f(CW\*(C`nss_delete\*(C'\fR call.  One workaround for this
bug seems to be to create add to the file \fI/etc/nsswitch.conf\fR
(at least) the following lines
.PP
.Vb 2
\&  group: files
\&  passwd: files
.Ve
.PP
Whether you are using \s-1NIS\s0 does not matter.  Amazingly enough,
the same bug also affects Solaris.
.SH "Miscellaneous"
.IX Header "Miscellaneous"
HP-UX 11 Y2K patch \*(L"Y2K\-1100 B.11.00.B0125 HP-UX Core \s-1OS\s0 Year 2000
Patch Bundle\*(R" has been reported to break the io/fs test #18 which
tests whether \fIutime()\fR can change timestamps.  The Y2K patch seems to
break \fIutime()\fR so that over \s-1NFS\s0 the timestamps do not get changed
(on local filesystems \fIutime()\fR still works). This has probably been
fixed on your system by now.
.SH "AUTHOR"
.IX Header "AUTHOR"
H.Merijn Brand <[e-mail]>
Jeff Okamoto <[e-mail]>
.PP
With much assistance regarding shared libraries from Marc Sabatella.
.SH "DATE"
.IX Header "DATE"
Version 0.8.0: 2007\-09\-09