Negligibly ( My Tech Blog )

Is it a 'char' or 'unsigned char' or 'signed char'?

Tuesday, 26 September 2006 17:18 administrator

I inherited a overly-loaded code something like this...

void func(boost::int8_t p) 
 {

 std::cout << "boost::int8_t" << std::endl;
 }
 void func(boost::uint8_t p)
 {
 std::cout << "boost::uint8_t" << std::endl;
 }
 void f1(boost::int16_t s)
 {
 std::cout << "boost::int16_t" << std::endl;
 }
 void f1(boost::uint16_t s)
 {
 std::cout << "boost::uint16_t" << std::endl;
 }
 main()
 {
 short sh;
 char c;
 f1(sh);
 func(c);
 }

I tried to compile the above code but the compiler is refusing to proceed balking at "func(c)", while it is completely happy with f1(sh). For the uninitiated boost types are typedef-ed to the appropriate platform types like int8_t -> signed char uint8_t -> unsigned char.

It turned out that the language does not specify whether variables of type char are signed or unsigned quantities. So the compiler complains about the func(c) call because it is ambiguous. Whereas f1(sh) succeeds.

Changind the "char c" to "unsigned char" or "signed char" resolves the ambiguity and calms down the compiler. Curiously even forcing the compiler with -funsigned-char didn't help in this case.

Well my philosophy of solving any problem by an extra level of indirection also didn't work. I tried something like this...


template<class> void func_char_resolver(T c);

template<> void func_char_resolver<boost::int8_t>( boost::int8_t c) 
{ 
 /// dealing with singed char....
}

template<> void func_char_resolver< boost::uint8_t >( boost::uint8_t c ) 
{ 
/// dealing with unsinged char....
}

void func( char c )
{
func_char_resolver< char >(c);
}

I don't know may be because `char' is a distinct type (of undefined sign) hence none of the specializations matched. Oh well!

Add new comment

Last Updated ( Saturday, 16 January 2010 03:40 )

Problems with X11 Composite Extension

Wednesday, 20 September 2006 22:46 administrator

When I enabled trancelucency in my KDE the composite manager crashed with the
message asking to insert the following lines in xorg.conf.

Section "Extensions"
Option "Composite" "Enable"
EndSection

And so I did and forgot about it for a week. Last week when I tried to launch
gnucash it wouldn't . It cried something like this.

-------
Gdk-CRITICAL **: file gdkwindow.c: line 1406 (gdk_window_get_visual):
assertion `window != NULL' failed.

Gdk-CRITICAL **: file gdkcolor.c: line 57 (gdk_colormap_new): assertion
`visual != NULL' failed.
SESSION_MANAGER=tcp/pdm1:53739

Gdk-CRITICAL **: file gdkwindow.c: line 1406 (gdk_window_get_visual):
assertion `window != NULL' failed.

Gdk-CRITICAL **: file gdkcolor.c: line 57 (gdk_colormap_new): assertion
`visual != NULL' failed.
Gdk-ERROR **: BadDrawable (invalid Pixmap or Window parameter)
serial 47 error_code 9 request_code 132 minor_code 5
Gdk-ERROR **: BadDrawable (invalid Pixmap or Window parameter)
serial 48 error_code 9 request_code 55 minor_code 0
--------

I thought one of yum update screwed it up. So I looked through the recent
update logs and couldn't find anything remotely connected to gnucash , X11,
gdk or my window manager. Frustrated like hell I stopped thinking about this
problem. I couldn't do my accounting for a week!

It wasn't until I tried to launch one more app which failed with the same
exception. I then started digging deep and found that the culprit was the
Composite extension option. So I disabled the following from xorg.conf

#Section "Extensions"
#Option "Composite" "Enable"
#EndSection

and yes I did my accounts immediately!

Add new comment

Kerberos on 64bit Linux with gcc 4.0

Friday, 08 September 2006 12:33 administrator

I was trying to compile kerberos on a 64 bit linux (CentOS). First I was getting configure errors about res_search(). Then I downloaded the latest and the greatest v5-1.5.1 from mit and tried to compile with gcc version 4.0. First the configure script was using the old compiler. So I need to teach it to use the gcc4. So I did...

export CC=gcc4
export CXX=g++4

After this configure script went fine. But the compilation was producing some errors like the following in the kadmin/test directory.

tcl_ovsec_kadm.c:85: `Tcl_HashEntry' undeclared (first use in this function)
tcl_ovsec_kadm.c:85: (Each undeclared identifier is reported only once
tcl_ovsec_kadm.c:85: for each function it appears in.)
tcl_ovsec_kadm.c:85: `entry' undeclared (first use in this function)
tcl_ovsec_kadm.c:93: warning: implicit declaration of function `Tcl_InitHashTable'
tcl_ovsec_kadm.c:93: `TCL_STRING_KEYS' undeclared (first use in this function)
tcl_ovsec_kadm.c:105: warning: implicit declaration of function`Tcl_CreateHashEntry'
tcl_ovsec_kadm.c:109: warning: implicit declaration of function `Tcl_SetHashValue'

Looks like I'm not alone -> http://mailman.mit.edu/pipermail/kerberos/2004-September/006391.html
So I followed it faitfully like this and everything went fine.

./configure --prefix=/home/test/krb5 --without-tcl

Add new comment

Getting the gcc/g++ compiler #define-s

Tuesday, 22 August 2006 13:35 administrator

I was doing 64 bit migration and wanted to know more about gcc/g++ compiler #define-s. One of my friend Homolka Richard gave me a very nice tip. Alias the whole thing!

alias whatg++='echo "main(){}" | g++ -E -x c++ -dM - '
alias whatgcc='echo "main(){}" | gcc -E -x c -dM - '

So running whatg++ from now on will list all the compiler #define-s something like this...

#define __HAVE_BUILTIN_SETJMP__ 1
#define __unix__ 1
#define unix 1
#define __i386__ 1
#define __SIZE_TYPE__ unsigned int
#define __ELF__ 1
#define __GNUC_PATCHLEVEL__ 3
#define __linux 1
#define __unix 1
#define __linux__ 1
#define __USER_LABEL_PREFIX__
#define linux 1
#define __STDC_HOSTED__ 1
#define __EXCEPTIONS 1
#define __GXX_WEAK__ 1
#define __WCHAR_TYPE__ long int
#define __gnu_linux__ 1
#define __WINT_TYPE__ unsigned int
#define __GNUC__ 3
#define __cplusplus 1
#define __DEPRECATED 1
#define __GNUG__ 3
#define __GXX_ABI_VERSION 102
#define i386 1
#define __GNUC_MINOR__ 2
#define __STDC__ 1
#define __PTRDIFF_TYPE__ int
#define __tune_i386__ 1
#define __REGISTER_PREFIX__
#define __NO_INLINE__ 1
#define _GNU_SOURCE 1
#define __i386 1
#define __VERSION__ "3.2.3 20030502 (Red Hat Linux 3.2.3-47.3)"

Add new comment

Last Updated ( Saturday, 16 January 2010 04:34 )

IMAP through SSL using pine

Wednesday, 16 August 2006 12:36 administrator

I have a unix/mail account on a external BSD server. Being a great fan of text mode I use fetchmail to retrieve the mails to my local linux machine and run elm or pine depending on my mood. But my mail server being too restictive on the smtp side, I couldn't hit reply on my local machine and send a reply or send a new mail. Relaying was denied. I could have used a different smtp server of course. But I wanted a simple solution.

So I setup pine on my local machine to do IMAP to my server and things are suddenly rosier. My local copy of pine works on the IMAP folders on my server. I only do fetchmail once in a week for archiving purposes.

The pine configuration is really simple...

Goto "Setup"->"Configuration"
Set 'inbox-path' to
'{mail.foo.com/ssl}INBOX'. Since my server requires a ssl connection todo IMAP.

If your server doesn't support ssl you can replace "ssl" with "notls". I also set the smtp-server to
mail.foo.com

Add new comment

More Articles...

<< Start < Prev 1 2 3 4 5 Next > End >>

Page 4 of 5