Howdy: To bring folks up to date, Dmalloc Library Version 4.0.3 was released on October 15. The 4.X.X releases contain significantly improved thread handling,...
Gray Watson
gray@...
Oct 26, 1998 3:52 pm
4
In my previous mail, I neglected to list the URL of the Debug Malloc library home page. http://www.letters.com/dmalloc/ Visit there to download the latest...
Gray Watson
gray@...
Oct 26, 1998 11:35 pm
5
Gray,
Please create a simple memalign() which simply prints a warning (on 1st invocation) and calls malloc() without regard to alignments. I have not seen a...
Robert D. Johnson
rjohnson@...
Nov 6, 1998 7:59 pm
6
We have a ton of code which initializes a ptr to NULL, then does free(ptr) later regardless of whether a malloc had been done. The documentation on Solaris...
Robert D. Johnson
rjohnson@...
Nov 6, 1998 8:04 pm
7
... Sorry Robert. I've always disagreed with this. I always do a: if (pnt != NULL) { free(pnt); } I think that freeing of NULLs is not good form and should...
Gray Watson
gray@...
Nov 9, 1998 3:51 pm
8
... First, I must say "thank you" for all your dmalloc work. I mean no disrespect. I am just trying help. ISO Standards discuss free(NULL) at: ...
Robert D. Johnson
RJohnson@...
Nov 9, 1998 5:37 pm
9
... I would say: 1) You can argue over whether the design is a good idea or not, however this is mostly a moot point. 2) You can assert that it's not portable...
Randell Jesup
rjesup@...
Nov 9, 1998 5:53 pm
10
... Good idea Robert. It has been completed as described above and is in my main source tree. A 4.1.0-beta version has been posted. Not too well tested. It...
Gray Watson
gray@...
Nov 9, 1998 7:46 pm
11
... Right. My guess is that memalign might have been more important 10 years ago, but perhaps not as important these days? I think it was used for speed,...
Robert D. Johnson
RJohnson@...
Nov 9, 1998 9:26 pm
12
... You haven't tried to use my system while I'm building a browser image... ;-) 128MB made it usable while doing the AR's and links; 64MB was unusable...
Randell Jesup
rjesup@...
Nov 10, 1998 2:04 am
13
So instead of an argument of semi-religious nature how about we utilize the egroups poll mechanisms. By default, should dmalloc allow by default the free (or ...
gwatson@...
Nov 10, 1998 5:35 pm
14
--w4C4UTlrpi Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Hi, dmalloc is a very fine package. However, I am offering the...
Jens Krinke
j.krinke@...
Nov 10, 1998 5:37 pm
15
... Regardless of the result, I think the documentation in settings.dist for free(0) needs to be changed to note that POSIX/ANSI/Unix/etc allow free(0)....
Randell Jesup
rjesup@...
Nov 10, 1998 7:36 pm
16
... `delete 0' is explicitly allowed by the C++ Standard, so, if free(0) is disallowed by dmalloc, operator delete should mask it. I don't care too much about...
Alexandre Oliva
oliva@...
Nov 10, 1998 8:05 pm
17
I still maintain that you should fix errant code. Due to the varying nature of runtimes and implementations of the C and various malloc libraries, you can't...
Scott Michel
scottm@...
Nov 10, 1998 8:46 pm
18
... Is this really true? Do we have a single real example of this, let alone enough to worry about? ... Hey, I hacked that game in Fortran (ok, so it wasn't...
Randell Jesup
rjesup@...
Nov 10, 1998 10:42 pm
19
... Careful - 0x00000000 being inaccessible and issues of free(0) are NOT linked. The whole point is that POSIX and all Unixes/PC compilers I know of...
Randell Jesup
rjesup@...
Nov 11, 1998 2:26 am
20
Hi everybody. I also found that free(NULL) is allowed by ANSI. But I belive it was included as a special case (with side effects) just to avoid invalidating...
Erick Recio
ferma@...
Nov 11, 1998 8:22 am
21
... AIX for one, at least the last time I ported to it. It's relatively common for the OS to make the page 0 PTE inaccessible. Used to be the case for Linux at...
Scott Michel
scottm@...
Nov 11, 1998 2:10 pm
22
... Hi everybody. I also found that free(NULL) is allowed by ANSI. But I belive it was included as a special case (with side effects) just to avoid...
Erick Recio
erecio@...
Nov 11, 1998 8:10 pm
23
... Hi everybody. I also found that free(NULL) is allowed by ANSI. But I belive it was included as a special case (with side effects) just to avoid...
Erick Recio
erecio@...
Nov 11, 1998 8:10 pm
24
After playing with dmalloc for a bit, I have a few suggestions: free-blank, check-blank: instead of blanking freed memory, set it to a pattern. alloc-blank:...
Randell Jesup
rjesup@...
Nov 12, 1998 12:32 am
25
I agree with some points made by Randell Jesup. I too would like user-defined patterns used in malloc and free areas. I too like the idea of splitting...
Robert D. Johnson
rjohnson@...
Nov 12, 1998 1:49 am
26
... I do. RTFM. It is set to all 0xc5. http://www.letters.com/dmalloc/texi/dmalloc_3.html#IDX163 ... free-zero and alloc-zero. Good idea. Will do. ... I...
Gray Watson
gray@...
Nov 12, 1998 2:41 pm
27
... How about 0xa110ca7e (allocate) and 0xf4eeda7a (freedata)? :-) -- Alexandre Oliva http://www.dcc.unicamp.br/~oliva aoliva@{acm.org} ...
Alexandre Oliva
oliva@...
Nov 12, 1998 2:41 pm
28
... Irrespective the what standards say, there are still buggy implementations which when confronted by free(0) will attempt to dereference page 0. So, in some...
Scott Michel
scottm@...
Nov 12, 1998 2:41 pm
29
... Cute. I've used patterns like that before - 0xC0DEDBAD, 0xDEADBEEF, 0xC0EDBABE (ok, that was a decade ago). Nowadays I prefer to use cute patterns like...
Randell Jesup
rjesup@...
Nov 12, 1998 7:25 pm
30
ov_seen_c and ov_iteration are int and long respectively. I've been seeing overflows(?) of ov_seen_c: 106944: not freed: '0x769f68|s87' (16 bytes) from...
Randell Jesup
rjesup@...
Nov 12, 1998 9:18 pm
31
... Most likely a dmalloc bug. I've seen them as well but I've never gotten a reproducable case that I can debug. Anyone got one for me? It would need to be...
Gray Watson
gray@...
Nov 12, 1998 10:11 pm
32
... You can't always extract this information. It would be not only platform-dependent; you'd have to rely on undocumented information about each particular...
Having problems with message search?
Fill out this form to ensure your group is one of the first to be migrated to the new message search system.
