Thread: question on some code.
Hi All,
I've been doing a code audit (review) for my own personal use of the
7.4.x series code base and have come across something that I'd like to get
more details on, if it is possible to do so.
I've been going over the communications section and am curious how this
segment of code is actually working:
/* -------------------------------- * pq_sendint - append a binary integer to a
StringInfo buffer * -------------------------------- */
void
pq_sendint(StringInfo buf, int i, int b)
{ unsigned char n8; uint16 n16; uint32 n32;
switch (b) { case 1: n8 = (unsigned char) i;
appendBinaryStringInfo(buf, (char *) &n8, 1); break; case 2:
n16 = htons((uint16) i); appendBinaryStringInfo(buf, (char *) &n16, 2);
break; case 4: n32 = htonl((uint32) i);
appendBinaryStringInfo(buf,(char *) &n32, 4); break; default:
elog(ERROR, "unsupported integer size %d", b); break; }
}
I understand the concept of the code, to append binary values to a string
buffer (char *), but, under my compiler on FreeBSD 5.4.x (gcc (GCC) 3.4.2
[FreeBSD] 20040728) I see a few issues that have cropped up.
If I understand the code right, your trying to pass in to
appendBinaryStringInfo an "address" or reference to the n8, n16, or n32
variables and cast them so that a char * pointer can access that address
space. Through some testing that I've been doing (outputting the values in
the appendBinaryStringInfo function), I never seem to see any data, The
variable pointer that references the "n8, n16, or n32" value is not holding
any data in the appendBinaryStringInfo function.
This might be my miss-understanding of the function and was hoping
someone would be willing to clearify it for me?
<blockquote type="CITE"><pre> <font color="#000000">If I understand the code right, your trying to pass in to </font> <font color="#000000">appendBinaryStringInfo an "address" or reference to the n8, n16, or n32 </font> <font color="#000000">variables and cast them so that a char * pointer can access that address </font> <font color="#000000">space. Through some testing that I've been doing (outputting the values in </font> <font color="#000000">the appendBinaryStringInfo function), I never seem to see any data, The </font> <font color="#000000">variable pointer that references the "n8, n16, or n32" value is not holding </font> <font color="#000000">any data in the appendBinaryStringInfo function.</font> </pre></blockquote><br /> How are you looking at the data? If you're treating the pointer as if it were a null-terminatedstring, it's likely that the first byte is a null character much of the time. Since this is binary data (notcharacter data), you'll need to look at the individual bytes. If you're using gdb, use 'x' command instead of the 'p'command. If you've thrown a printf() into the code, you can't use a "%s" format specifier, you'll have to look at thebytes one at a time using something like "%02X".<br /><br /><br /> -- Korry
Chris Bowlby <excalibur@accesswave.ca> writes:
> I understand the concept of the code, to append binary values to a string
> buffer (char *), but, under my compiler on FreeBSD 5.4.x (gcc (GCC) 3.4.2
> [FreeBSD] 20040728) I see a few issues that have cropped up.
You've got a broken compiler then, because it certainly works for
everyone else. That function's been substantially unchanged since
... hmm ... version 1.1 of pqformat.c dated 25 Apr 1999. And I'm
fairly sure that I got it from some older code rather than writing
it from scratch when I made that file.
> Through some testing that I've been doing (outputting the values in
> the appendBinaryStringInfo function), I never seem to see any data, The
> variable pointer that references the "n8, n16, or n32" value is not holding
> any data in the appendBinaryStringInfo function.
Define "never seem to see any data"...
regards, tom lane
Ahh, so it's not attempting to turn it into an "acsii" string, just storing raw binary data... Ok that makes sense now, thanks for the help. At 02:54 PM 7/19/2005, Korry wrote: >>If I understand the code right, your trying to pass in to >>appendBinaryStringInfo an "address" or reference to the n8, n16, or n32 >>variables and cast them so that a char * pointer can access that address >>space. Through some testing that I've been doing (outputting the values in >>the appendBinaryStringInfo function), I never seem to see any data, The >>variable pointer that references the "n8, n16, or n32" value is not holding >>any data in the appendBinaryStringInfo function. > >How are you looking at the data? If you're treating the pointer as if it >were a null-terminated string, it's likely that the first byte is a null >character much of the time. Since this is binary data (not character >data), you'll need to look at the individual bytes. If you're using gdb, >use 'x' command instead of the 'p' command. If you've thrown a printf() >into the code, you can't use a "%s" format specifier, you'll have to look >at the bytes one at a time using something like "%02X". > > > -- Korry
Hi Tom, good to hear from you. It was my miss-understanding of what the code was doing, not the issue with the compiler :> At 03:22 PM 7/19/2005, Tom Lane wrote: >Chris Bowlby <excalibur@accesswave.ca> writes: > > I understand the concept of the code, to append binary values to a string > > buffer (char *), but, under my compiler on FreeBSD 5.4.x (gcc (GCC) 3.4.2 > > [FreeBSD] 20040728) I see a few issues that have cropped up. > >You've got a broken compiler then, because it certainly works for >everyone else. That function's been substantially unchanged since >... hmm ... version 1.1 of pqformat.c dated 25 Apr 1999. And I'm >fairly sure that I got it from some older code rather than writing >it from scratch when I made that file. > > > Through some testing that I've been doing (outputting the values in > > the appendBinaryStringInfo function), I never seem to see any data, The > > variable pointer that references the "n8, n16, or n32" value is not > holding > > any data in the appendBinaryStringInfo function. > >Define "never seem to see any data"... > > regards, tom lane