Bug#326451: jlint: FTBFS (amd64): cast from 'const void*' to 'int' loses precision
On Fri, Sep 09, 2005 at 09:16:04AM +0530, Ganesan Rajagopal wrote: Kurt Roeckx wrote: It looks like the patch was only partially applied? Anyway, please do not cast a pointer to either an int or a long. Cast it to an intptr_t if you have to, but you should be able to avoid that too. Thanks. But how do you give a printf format argument for intptr_t? You shouldn't. You passed an integer to it, so extract an integer from the va_list. You could cast it back to an int if you want to. If you really want to print an intptr_t, inttypes.h defines a PRIdPTR you can use to print an intptr_t, something like: printf(% PRIdPTR \n, ptr); Kurt -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]
Bug#326451: jlint: FTBFS (amd64): cast from 'const void*' to 'int' loses precision
Kurt Roeckx wrote: On Fri, Sep 09, 2005 at 09:16:04AM +0530, Ganesan Rajagopal wrote: Thanks. But how do you give a printf format argument for intptr_t? You shouldn't. You passed an integer to it, so extract an integer from the va_list. You could cast it back to an int if you want to. Oh we already agreed about that :-). I was curious how to handle intptr_t portably in a printf. Thanks for the info. Ganesan -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]
Bug#326451: jlint: FTBFS (amd64): cast from 'const void*' to 'int' loses precision
found 326451 3.0-3 thanks Hi, 3.0-3 is still failing to build: jlint.cc: In function 'void format_message(int, const utf_string, int, __va_list_tag*)': jlint.cc:240: warning: field precision should have type 'int', but argument 3 has type 'long int' jlint.cc:264: error: cast from 'const void*' to 'int' loses precision It looks like the patch was only partially applied? Anyway, please do not cast a pointer to either an int or a long. Cast it to an intptr_t if you have to, but you should be able to avoid that too. It seems to be using stdargs. It really should just call va_arg() with the right type instead of always using a void *. If you always call va_arg for a void *, this means whenever you pass something to it, you'll always have to cast it to a void *. PS: The TODO file says: * (1) Make code 64 bit clean by adding more type casts. E.g. void *p; int i; p = (void *) (ptrword) i; where ptrword is derived from a ./configure script. ptrword should just be intptr_t. And he shouldn't be using it. PS: It would be nice if you could run the testsuite (in tests.tar.bz2) and make the build fail if it found any errors. Kurt -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]
Bug#326451: jlint: FTBFS (amd64): cast from 'const void*' to 'int' loses precision
Kurt Roeckx wrote: found 326451 3.0-3 thanks I had noted the problem myself and already uploaded 3.0-4. 3.0-4 has already been autobuilt successfully for ia64 and alpha. http://buildd.debian.org/build.php?arch=pkg=jlint It looks like the patch was only partially applied? Anyway, please do not cast a pointer to either an int or a long. Cast it to an intptr_t if you have to, but you should be able to avoid that too. Thanks. But how do you give a printf format argument for intptr_t? It seems to be using stdargs. It really should just call va_arg() with the right type instead of always using a void *. If you always call va_arg for a void *, this means whenever you pass something to it, you'll always have to cast it to a void *. I agree. I have no idea why upstream took this approach. PS: It would be nice if you could run the testsuite (in tests.tar.bz2) and make the build fail if it found any errors. I'll take a look. Ganesan -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]
Bug#326451: jlint: FTBFS (amd64): cast from 'const void*' to 'int' loses precision
reopen 326451
retitle 326451 FTBFS (amd64): cast from 'const void*' to 'int' loses precision
thanks
Hello,
thanks for fixing the Build-Depends.
However, the patch contained a second part which was not applied.
When building 'jlint' on amd64/unstable, I still get the following error:
jlint.cc: In function 'void format_message(int, const utf_string, int,
__va_list_tag*)':
jlint.cc:153: error: cast from 'const void*' to 'int' loses precision
jlint.cc:228: error: cast from 'const void*' to 'int' loses precision
jlint.cc:240: warning: field precision should have type 'int', but argument 3
has type 'long int'
jlint.cc:264: error: cast from 'const void*' to 'int' loses precision
jlint.cc: In function 'bool parse_class_file(byte*)':
jlint.cc:431: warning: unused variable 'desc_index'
jlint.cc:496: warning: unused variable 'max_stack'
jlint.cc:325: warning: unused variable 'minor_version'
jlint.cc:327: warning: unused variable 'major_version'
make[1]: *** [jlint.o] Error 1
make[1]: Leaving directory `/jlint-3.0'
make: *** [build-arch-stamp] Error 2
With the attached patch 'jlint' can be compiled
on amd64 using gcc-4.0.
Regards
Andreas Jochens
diff -urN ../tmp-orig/jlint-3.0/jlint.cc ./jlint.cc
--- ../tmp-orig/jlint-3.0/jlint.cc 2004-01-14 16:23:20.0 +
+++ ./jlint.cc 2005-09-03 11:38:32.0 +
@@ -137,7 +137,7 @@
void format_message(int code, utf_string const file, int line, __VALIST ap)
{
- static int loop_id;
+ static long loop_id;
static message_node *first, *last;
static char* compound_message;
const void* parameter[MAX_MSG_PARAMETERS];
@@ -150,7 +150,7 @@
if (compound_message != NULL
((loop_id != 0
((code != msg_loop code != msg_sync_loop)
- || (int)parameter[2] != loop_id))
+ || (long)parameter[2] != loop_id))
|| (loop_id == 0 code != msg_wait_path)))
{
if (!message_node::find(compound_message)) {
@@ -225,7 +225,7 @@
name.as_asciz());
break;
case 'd': // integer
-dst += sprintf(dst, %d, (int)parameter[index]);
+dst += sprintf(dst, %ld, (long)parameter[index]);
break;
default:
assert(false/*bad message parameter format*/);
@@ -261,7 +261,7 @@
compound_message = strdup(his_buf);
first = last = new message_node(msg_buf);
if (code != msg_wait) {
-loop_id = (int)parameter[2];
+loop_id = (long)parameter[2];
}
} else if (!message_node::find(his_buf)) {
fprintf(stdout, %s\n, msg_buf);
--
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]
