November 2007 - Darwin-dev - lists.apple.com

Re: fcntl questions
by Steve Checkoway 04 Nov '07

04 Nov '07

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com On Nov 4, 2007, at 7:32 AM, Thierry Passeron wrote: -- Steve Checkoway _______________________________________________ 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 The reason why they use SIGRTMIN + x in the linux example is that they use the possibility to get some si_fd extra member in the siginfo_t struct passed as argument to the signal handler installed with sigaction(). And so it is possible to know which descriptor became ready when the signal was sent (no need to use select() in the signal handler) I have not found this si_fd member in the Darwin siginfo_t structure. So how can I proceed ? I know I could use the default signals SIGIO/ SIGURG and use a select() in the signal handler to check which descriptor are available for reading now. I can't answer all of your questions but the only functions you should call in a signal handler are those listed in the sigaction man page. select(2) is not among them. If I were doing this and I really needed to not block in select, I'd either catch the SIGIO signal and set a flag (which is about all you should ever do in a signal handler) which I would check each time through a loop and when it is set, I'd use select with a zero timeout to get the pipes that need to be read from and then read from them. Of course, I don't know how often one gets that signal. I would think it would be easier to either use poll(2) or select(2) each time through the loop to be sure I was catching all of the data that was ready to read. smime.p7s

1 0

fcntl questions
by Thierry Passeron 04 Nov '07

04 Nov '07

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com Dkim-signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=beta; h=domainkey-signature:received:received:message-id:from:to:in-reply-to:content-type:content-transfer-encoding:mime-version:subject:date:references:x-mailer; bh=ylJ9NA+csxGTsKvrZj6YvF4icB1vjIx17vYUuVEwLJQ=; b=hUFhmgp2Cy6O+iPEX+R9JFmHwkUu7ViTLcx+LOAOQW7Gbf8rdITJF6myfhhl8k912Fs/IZlNLYpEP92sgvOaZkwdoiAJY+DEtUFfw5AZJ47EVCh3ikZJBAe4YU8gs9opvLW6AsoUHQvZRG44p78mAULYQgzWNoZsDvWpKqC6g58= Domainkey-signature: a=rsa-sha1; c=nofws; d=gmail.com; s=beta; h=received:message-id:from:to:in-reply-to:content-type:content-transfer-encoding:mime-version:subject:date:references:x-mailer; b=rzfXchb05otOOkT8IygIQbER+KK9noTkUXcPBE9FGTgbdAtEmBTFnpaJYilH+NQpcn64WdoXxAEaAVCGhPCSt7lcgJwuplPZLyyB7G3Bueg/AgKt4WTVAX7jki1RdY3m9vnl2wMfnubid7kdSMcfeg/3tM6HYPI5rlpxDcO99SQ= Hello people, * (SIGRTMIN + x) signals as callback functions for fcntl() Any advise is welcome. Regards, Thierry _______________________________________________ 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 writting a little program in C for educational and personal use in which I need to read from 2 descriptors piped from the stdout and stderr of an other forked process. I already figured out how to do all the pipings and the descriptors closing in both father and son processes. Also I have already done the reading of ready descriptors returned by select(). But now I'd like to do the same with fcntl() so that the process is not blocked on select() and can do anything else when the descriptors are not ready to be read. I have read some documentation about fcntl and some linux examples, so I tried to adapt them to Darwin (on OSX 10.5) and could not compile because of some errors. The example uses: Is it possible to set a different (other than SIGIO or SIGURG) signal for fcntl callbacks ? If yes, I have not found in the fcntl man page any reference to the F_SETSIG function ... Does Darwin allow you to have SIGRTMIN, SIGRTMAX ? I have not seen it in the output of /bin/kill -l. Forgive my ignorance but is it possible to send a non existant signal like (SIGUSR2 + x) ? The reason why they use SIGRTMIN + x in the linux example is that they use the possibility to get some si_fd extra member in the siginfo_t struct passed as argument to the signal handler installed with sigaction(). And so it is possible to know which descriptor became ready when the signal was sent (no need to use select() in the signal handler) I have not found this si_fd member in the Darwin siginfo_t structure. So how can I proceed ? I know I could use the default signals SIGIO/ SIGURG and use a select() in the signal handler to check which descriptor are available for reading now. This email sent to site_archiver(a)lists.apple.com

1 0

Re: gcc linking question
by Steve Checkoway 03 Nov '07

03 Nov '07

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com On Nov 3, 2007, at 10:03 AM, mothra wrote: So in your example, this should be fine: // Merp.h class Merp { // ... static const double Thing = 1.0; }; // Merp.cc const double Merp::Thing; That's probably more akin to what you want. -- Steve Checkoway _______________________________________________ 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 Excellent. Apple is not a rogue entity, my code is now more compliant, *and* I now understand how the compiler was handling my old code. Thanks so much for the help! It might be worth noting this part of the standard (yes, I was too tired to look at it last night, but not now =), 9.4.2.4 [class.static.data]: If a static data member is of const integral or const enumeration type, its declaration in the class definition can specify a constant- initializer which shall be an integral constant expression within its scope. The member shall still be defined in a namespace scope if it is used in the program and the namespace scope definition shall not contain an initializer. smime.p7s

1 0

Re: gcc linking question
by mothra 03 Nov '07

03 Nov '07

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com User-agent: Mutt/1.4.2.1i Excellent. Apple is not a rogue entity, my code is now more compliant, *and* I now understand how the compiler was handling my old code. Thanks so much for the help! - Lowell On Sat, Nov 03, 2007 at 08:52:48AM -0700, Duane Murphy wrote: > --- At Sat, 3 Nov 2007 09:21:09 -0500, mothra wrote: > > > > >Thanks for the input - it appears that that solution does work. It is a > >bit distressing to me that Apple's gcc 4.0.1 is not consistent with either > >RedHat's gcc 4.1.* or their gcc 3.*, or Apple's gcc 3.* in this regard. Is > >Apple striking out on its own here? Or is this requirement for two lines > >of code to initialize such variables an existing standard that only Apple > >is currently enforcing? > > More recent compilers have changed. > > I used to do exactly what you did. Previous compilers would essentially > treat the initialization as a immediate constant. That is the actual > variable and memory would not be allocated, but just use the initializer > value everywhere. > > There must have been a determination that this was technically not > allowed. The solution is to do as the standard says and specify the > initializer separately. This has nothing to do with Apple, but rather > the maturation of the compilers. > > ...Duane _______________________________________________ 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: gcc linking question
by Duane Murphy 03 Nov '07

03 Nov '07

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com --- At Sat, 3 Nov 2007 09:21:09 -0500, mothra wrote: > >Thanks for the input - it appears that that solution does work. It is a >bit distressing to me that Apple's gcc 4.0.1 is not consistent with either >RedHat's gcc 4.1.* or their gcc 3.*, or Apple's gcc 3.* in this regard. Is >Apple striking out on its own here? Or is this requirement for two lines >of code to initialize such variables an existing standard that only Apple >is currently enforcing? More recent compilers have changed. I used to do exactly what you did. Previous compilers would essentially treat the initialization as a immediate constant. That is the actual variable and memory would not be allocated, but just use the initializer value everywhere. There must have been a determination that this was technically not allowed. The solution is to do as the standard says and specify the initializer separately. This has nothing to do with Apple, but rather the maturation of the compilers. ...Duane _______________________________________________ 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: gcc linking question
by Shawn Erickson 03 Nov '07

03 Nov '07

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com On Nov 3, 2007, at 7:21 AM, mothra wrote: Versions of gcc before 4.0 had much loser standards compliance. -Shawn _______________________________________________ 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… Thanks for the input - it appears that that solution does work. It is a bit distressing to me that Apple's gcc 4.0.1 is not consistent with either RedHat's gcc 4.1.* or their gcc 3.*, or Apple's gcc 3.* in this regard. Is Apple striking out on its own here? Or is this requirement for two lines of code to initialize such variables an existing standard that only Apple is currently enforcing? Without looking at the compile command lines for your RH compile it is hard to say why it is working. I am fairly curtain that what you had isn't allowed by the standard and so Apple gcc 4.0.1 is being compliant. RH gcc 4.1 should also be compliant but you may be specifying a mode that allows some "gcc" specific extension to the standard. This email sent to site_archiver(a)lists.apple.com

1 0

Re: gcc linking question
by mothra 03 Nov '07

03 Nov '07

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com User-agent: Mutt/1.4.2.1i Thanks for the input - it appears that that solution does work. It is a bit distressing to me that Apple's gcc 4.0.1 is not consistent with either RedHat's gcc 4.1.* or their gcc 3.*, or Apple's gcc 3.* in this regard. Is Apple striking out on its own here? Or is this requirement for two lines of code to initialize such variables an existing standard that only Apple is currently enforcing? - Lowell _______________________________________________ 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: gcc linking question
by Steve Checkoway 03 Nov '07

03 Nov '07

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com On Nov 2, 2007, at 2:09 PM, mothra wrote: // In Merp.h class Merp { // ... static const double Thing; }; // In Merp.cpp const double Merp::Thing = 1.0; -- Steve Checkoway _______________________________________________ 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 I get lip from the linker when I do this - it claims that Thing is an undefined symbol. Is there a trick to assigning values to variables within the class declaration with the gcc that ships with Xcode 2.4 etc? Thanks for any help! I believe that the linker is correct here. For once, I don't feel like digging through the standard, but you declare the static variables in the class declaration and then in some source file, you define it. E.g., smime.p7s

1 0

gcc linking question
by mothra 02 Nov '07

02 Nov '07

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com User-agent: Mutt/1.4.2.1i Hi all, Have searched around for clues to this, but to no avail. The following code used to compile with older versions of gcc (eg, with Panther). It still compiles with gcc 4.1.1 on my Fedora system, but appears broken on the current Xcode for 10.4: -------------------------------------------------------------------------------- #include <iostream> class Merp{ public: void Print(){std::cout << Merp::Thing << std::endl;} private: const static double Thing = 1.00; }; int main(){ Merp asdf; asdf.Print(); } -------------------------------------------------------------------------------- I get lip from the linker when I do this - it claims that Thing is an undefined symbol. Is there a trick to assigning values to variables within the class declaration with the gcc that ships with Xcode 2.4 etc? Thanks for any help! - Lowell _______________________________________________ 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: CFRunLoopStop too powerful when invoked by bootstrap per-user daemon?
by Pete Gontier 02 Nov '07

02 Nov '07

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com Thread-index: Acgdbu7LLXGCJIliEdy+/gARJG7C6AAA1ywEAATZRHA= Thread-topic: CFRunLoopStop too powerful when invoked by bootstrap per-user daemon? User-agent: Microsoft-Entourage/11.3.6.070618 _______________________________________________ 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