August 2005 - Darwin-dev - lists.apple.com

Link Error -- Relocation Overflow [SOLVED]
by Bob Clark 23 Aug '05

23 Aug '05

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com My original message is at: http://lists.apple.com/archives/Darwin-dev/2005/Jun/msg00120.html and my followup is at: http://lists.apple.com/archives/Darwin-dev/2005/Jul/msg00044.html How to avoid the problem: using -mlong-branch can help. 1: ... bne 1 And the fix (!) was to change it to look like this: 1: ... bne 1b --Bob _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-dev mailing list (Darwin-dev(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-dev/site_archiver%40lists.app… I ended up blowing a DTS incident and George Warner provided the help that led to a solution. For the benefit of Googlers Of The Future, I'm providing a synopsis of what DTS suggested, and a description of what turned out to solve my problem. What the error message means: the PowerPC unconditional branch is 24 bits and the conditional branches are 14 bits. How to diagnose more fully: extract the .o from the .a file using the ar command, then run otool -rv on the .o file. Also run nm -n on the .o file to match up the symbols. In my case I had a conditional branch that was trying to branch over more than 14 bits, and this branch was in raw assembly code. (If it was in C code, then the -mlong-branch flag would probably have helped, but since it was in raw assembly the bne wasn't translated to avoid the 14-bit branch.) Specifically, I had an inline assembly function that included code like this: Adding the "b" is a compiler hint to branch backwards. Without the "b", it seems that it was trying to branch to a different invocation of this (inline) function, which was a long long ways away. With the "b" it branches backwards to the correct label. Anyway, dunno if this is of interest to anyone, but if there's even a chance that this will help ease the next guy's frustration it'll be worth it. This email sent to site_archiver(a)lists.apple.com

1 0

Re: Link problem on download site
by Kevin Van Vechten 19 Aug '05

19 Aug '05

site_archiver(a)lists.apple.com Delivered-To: Darwin-dev(a)lists.apple.com Hi Ben, Thanks for pointing this out. It's fixed now. - Kevin On Aug 18, 2005, at 6:42 PM, Benjamin Herrenschmidt wrote: Hi ! Regards, Ben. _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-dev mailing list (Darwin-dev(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-dev/site_archiver%40lists.app… The ".tar.gz" download link for AppleMacRISC4PE in Darwin 8.2 brings you back to the main page directly instead of downloading anything. There is no problem with accessing individual files via HTML interface though. This email sent to site_archiver(a)lists.apple.com

1 0

lm
by plumber 19 Aug '05

19 Aug '05

site_archiver(a)lists.apple.com Delivered-To: Darwin-dev(a)lists.apple.com i've built lm lmx for darwin 7 x86 if someone is interested it's here I did not have much time to answer and read my emails these last weeks I thank Jordan for the answers more developed a than mine I am sincerely sorry for my short answers but I give tracks :) ps : the link on the old page of dev it was a joke :) Best Regards -----BEGIN PLUMBER CODE BLOCK----- Version: 1.0 ¬ | | > Regards Plumber || GNU Darwin. ¬ \\\ plumber.gnu-darwin.org ¬ ¬ | | > OOP --\///\ (0)=(0) Darwin/Power Mac G4. ¬ | | > ( __ò ó__ ) The best for Fede ¬ \\\ federicafontana.it ¬ ------END PLUMBER CODE BLOCK------ -----BEGIN GEEK CODE BLOCK----- Version: 3.1 GMU/S d+@ s: a- K+++++ PGP++++ K---- dpu s--:-- C++++ C-- B L X U++++ P+++ C-- P! L--- E+++ W+++ P! L--- N+++ o-- K+++++ ------END GEEK CODE BLOCK------ Plain, Loving, Useful, Mischievous, Brutal, Explosive, Responsible _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-dev mailing list (Darwin-dev(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-dev/site_archiver%40lists.app… I"It's making my eyes better. And it's legal. I could walk up to the President and blow smoke in his stupid monkey face, and he'd just have to sit there groovin' on it!" Homer J Simpson, Springfield, Illinois, Massachusetts nobody knows :) This email sent to site_archiver(a)lists.apple.com

1 0

Link problem on download site
by Benjamin Herrenschmidt 19 Aug '05

19 Aug '05

site_archiver(a)lists.apple.com Delivered-To: Darwin-dev(a)lists.apple.com Hi ! The ".tar.gz" download link for AppleMacRISC4PE in Darwin 8.2 brings you back to the main page directly instead of downloading anything. There is no problem with accessing individual files via HTML interface though. Regards, Ben. _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-dev mailing list (Darwin-dev(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-dev/site_archiver%40lists.app… This email sent to site_archiver(a)lists.apple.com

1 0

Re: directio, O_DIRECT, "mount -o sync", O_SYNC, O_DSYNC and all that
by Sam Vaughan 18 Aug '05

18 Aug '05

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com On 19/08/2005, at 5:05 AM, Reto Kramer <kramer(a)acm.org> wrote: Hi Reto, Try calling fcntl() with F_NOCACHE to get Direct I/O behaviour. Sam _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-dev mailing list (Darwin-dev(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-dev/site_archiver%40lists.app… I'm looking for a way to (1) enable direct IO on OS X (10.4) This email sent to site_archiver(a)lists.apple.com

1 0

Re: core dump files
by Hugh Hoover 18 Aug '05

18 Aug '05

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com On Aug 18, 2005, at 09:06, Dave Zarzycki wrote: That no longer works in Tiger as you discovered. Do this now: echo "limit core unlimited" >> /etc/launchd.conf reboot Thanks! - that worked... _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-dev mailing list (Darwin-dev(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-dev/site_archiver%40lists.app… This email sent to site_archiver(a)lists.apple.com

1 0

Re: dlopen and d2i_DHparams
by Steve Checkoway 18 Aug '05

18 Aug '05

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com On Aug 18, 2005, at 1:45 PM, Kevin Van Vechten wrote: _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-dev mailing list (Darwin-dev(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-dev/site_archiver%40lists.app… This email sent to site_archiver(a)lists.apple.com What is the difference between SHA1 and d2i_DHparams that would cause the first program to fail but the second one to succeed? The obvious difference in the code you provided is that SHA1 is function invocation and d2i_DHparams is a function reference. I don't know enough about dlopen semantics to know whether the reference should be resolved or not, but I'm sure you could use dlsym(3) to get a valid function reference. That is what I ended up doing. I also noticed that exported variables behave in the same way as d2i_DHparams (and i2d_DHparams for that matter) did in my example. - Steve smime.p7s

1 0

Re: dlopen and d2i_DHparams
by Kevin Van Vechten 18 Aug '05

18 Aug '05

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com - Kevin On Aug 17, 2005, at 12:44 AM, Steve Checkoway wrote: $ cat foobar.cpp #include <openssl/dh.h> #include <openssl/asn1.h> #include <openssl/bn.h> #include <cstdio> #include <dlfcn.h> DH *dh2 = DHparams_dup(dh); dlclose(handle); return 0; } $ make foo bar baz g++ -o foo -lcrypto foobar.cpp g++ -flat_namespace -undefined suppress -bind_at_load \ -o bar foobar.cpp MACOSX_DEPLOYMENT_TARGET=10.3 g++ -undefined dynamic_lookup \ -bind_at_load -o baz foobar.cpp $ ./foo $ ./bar dyld: Symbol not found: _d2i_DHparams Referenced from: /Users/steve/temp/dlopen/./bar Expected in: flat namespace Trace/BPT trap $ ./baz dyld: Symbol not found: _d2i_DHparams Referenced from: /Users/steve/temp/dlopen/./baz Expected in: dynamic lookup Trace/BPT trap $ cat hash.cpp #include <openssl/sha.h> #include <cstdio> #include <cstring> #include <dlfcn.h> puts(hello); SHA1((unsigned char *)hello, strlen(hello), md); for(int i = 0; i < SHA_DIGEST_LENGTH; ++i) printf("%02hhx", md[i]); puts(""); dlclose(handle); return 0; } $ make hash g++ -flat_namespace -undefined suppress -bind_at_load \ -o hash hash.cpp $ ./hash Hello World! 2ef7bde608ce5404e97d5f042f95f89f1c232871 This email sent to kevin(a)opendarwin.org _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-dev mailing list (Darwin-dev(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-dev/site_archiver%40lists.app… What is the difference between SHA1 and d2i_DHparams that would cause the first program to fail but the second one to succeed? The obvious difference in the code you provided is that SHA1 is function invocation and d2i_DHparams is a function reference. I don't know enough about dlopen semantics to know whether the reference should be resolved or not, but I'm sure you could use dlsym (3) to get a valid function reference. I'm getting strange behavior when I try to dlopen libcrypto.dylib and then use d2i_DHparams. #ifdef DHparams_dup # undef DHparams_dup #endif #define DHparams_dup (x) \ (DH *)ASN1_dup((int (*)(...)) i2d_DHparams, \ (char *(*)(...))d2i_DHparams, (char *)(x)) int main() { void *handle = dlopen("libcrypto.dylib", RTLD_LAZY| RTLD_GLOBAL); DH *dh = DH_new(); BN_hex2bn(&dh->p, "FFFFFFFFFFFFFFFFC90FDAA22168C234C4C6628B80DC1CD129024" "E088A67CC74020BBEA63B139B22514A08798E3404DDEF9519B3CD" "3A431B302B0A6DF25F14374FE1356D6D51C245E485B576625E7EC" "6F44C42E9A637ED6B0BFF5CB6F406B7EDEE386BFB5A899FA5AE9F" "24117C4B1FE649286651ECE65381FFFFFFFFFFFFFFFF"); dh->g = BN_new(); BN_set_word(dh->g, DH_GENERATOR_2); When I run ./foo it runs correctly. When I run ./bar and ./baz dyld fails with the symbol not found. However, it works with other functions such as SHA1. int main() { void *handle = dlopen("libcrypto.dylib", RTLD_LAZY| RTLD_GLOBAL); unsigned char md[SHA_DIGEST_LENGTH]; char hello[] = "Hello World!"; What is the difference between SHA1 and d2i_DHparams that would cause the first program to fail but the second one to succeed? - Steve _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-dev mailing list (Darwin-dev(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-dev/kevin% 40opendarwin.org This email sent to site_archiver(a)lists.apple.com

1 0

Re: core dump files
by Dave Zarzycki 18 Aug '05

18 Aug '05

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com That no longer works in Tiger as you discovered. Do this now: echo "limit core unlimited" >> /etc/launchd.conf reboot On Aug 18, 2005, at 9:03 AM, Hugh Hoover wrote: When I check ulimit -a, I get: core file size (blocks, -c) 0 ... but when I attempt to change it, i.e.: > ulimit -c unliimited > ulimit -c 0 It remains stubbornly at zero. I"ve also tried 1024 instead of unlimited, no difference. Does anyone have some insight into this? Any suggestions? _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-dev mailing list (Darwin-dev(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-dev/zarzycki%40apple.com This email sent to zarzycki(a)apple.com _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-dev mailing list (Darwin-dev(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-dev/site_archiver%40lists.app… Well, I can't get a core file generated to save my life! I have COREDUMPS=-YES- in my hostconfig, and the /cores directory is world-writeable. This email sent to site_archiver(a)lists.apple.com

1 0

core dump files
by Hugh Hoover 18 Aug '05

18 Aug '05

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com When I check ulimit -a, I get: core file size (blocks, -c) 0 ... but when I attempt to change it, i.e.: > ulimit -c unliimited > ulimit -c 0 It remains stubbornly at zero. I"ve also tried 1024 instead of unlimited, no difference. Does anyone have some insight into this? Any suggestions? _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-dev mailing list (Darwin-dev(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-dev/site_archiver%40lists.app… Well, I can't get a core file generated to save my life! I have COREDUMPS=-YES- in my hostconfig, and the /cores directory is world-writeable. This email sent to site_archiver(a)lists.apple.com

1 0