Bug#318525: FTBFS: Memory inputs not directly addressable
Hi Roger, On Sun, Jul 31, 2005 at 05:22:15PM +0100, Roger Leigh wrote: On Sun, Jul 31, 2005 at 04:54:14PM +0100, Roger Leigh wrote: /tmp/buildd/atlas-3.2.1ln/include/contrib/camm_dpa.h:766: error: memory input 2 is not directly addressable I've been able to compile this file with the attached patch. However, I have some troubles building the package: The first debuild -uc -us fails with (before any compilation): Looking for BLAS (this may take a while): make: *** [bin/Linux_3dnow_static/Make.inc] Error 1 debuild: fatal error at line 765: dpkg-buildpackage failed! Then, a second debuild -uc -us compiles some files, then install libblas.a in the debian/tmp directory, and then I have the same failure: Looking for BLAS (this may take a while): make: *** [bin/Linux_3dnow_static/Make.inc] Error 1 debuild: fatal error at line 765: dpkg-buildpackage failed! Any idea? hth, -- Nekral --- /tmp/camm_dpa.h 2005-08-06 22:12:52.0 +0200 +++ include/contrib/camm_dpa.h 2005-08-06 22:27:38.0 +0200 @@ -762,6 +762,9 @@ #define fixm c #endif NO_INLINE; +int stride_size = stride*sizeof(*fixm); +int lda_size = lda*sizeof(*a); +int len_size = len*sizeof(*movm)/sizeof(float); ASM ( @@ -884,8 +887,8 @@ popl %%ebx\n\t -::m (movm),m (fixm),m (stride*sizeof(*fixm)),m (a),m (lda*sizeof(*a)), - m (len*sizeof(*movm)/sizeof(float)) +::m (movm),m (fixm),m (stride_size),m (a),m (lda_size), + m (len_size) #ifdef SCPLX ,m (w)
Bug#318525: FTBFS: Memory inputs not directly addressable
-BEGIN PGP SIGNED MESSAGE- Hash: SHA1 To follow up: /tmp/buildd/atlas-3.2.1ln/include/contrib/camm_dpa.h: In function 'dp2dpNb0': /tmp/buildd/atlas-3.2.1ln/include/contrib/camm_dpa.h:766: error: memory input 2 is not directly addressable /tmp/buildd/atlas-3.2.1ln/include/contrib/camm_dpa.h:766: error: memory input 4 is not directly addressable /tmp/buildd/atlas-3.2.1ln/include/contrib/camm_dpa.h:766: error: memory input 5 is not directly addressable Line 766 is the start of a block of ix86 inline assembly. Unfortunately, I do not know assembly. The ASM macro expands to __asm__ __volatile__, but GCC 4.0 does not document the __volatile__ keyword. Changing it to volatile might help, but I don't think the error lies at this point. Regards, Roger - -- Roger Leigh Printing on GNU/Linux? http://gimp-print.sourceforge.net/ Debian GNU/Linuxhttp://www.debian.org/ GPG Public Key: 0x25BFB848. Please sign and encrypt your mail. -BEGIN PGP SIGNATURE- Version: GnuPG v1.4.1 (GNU/Linux) Comment: Processed by Mailcrypt 3.5.8 http://mailcrypt.sourceforge.net/ iD8DBQFC7PQlVcFcaSW/uEgRAssKAKCBd/Zh+3ZVnsZQfbKVEJABvtJwsgCcDMDI FB1NQsmvqjYa9RYPOXOOaIE= =d6Km -END PGP SIGNATURE- -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]
Bug#318525: FTBFS: Memory inputs not directly addressable
On Sun, Jul 31, 2005 at 04:54:14PM +0100, Roger Leigh wrote: /tmp/buildd/atlas-3.2.1ln/include/contrib/camm_dpa.h:766: error: memory input 2 is not directly addressable Line 766 is the start of a block of ix86 inline assembly. Unfortunately, I do not know assembly. The ASM macro expands to __asm__ __volatile__, but GCC 4.0 does not document the __volatile__ keyword. Changing it to volatile might help, but I don't think the error lies at this point. Changing the VOLATILE macro from __volatile__ to volatile makes no difference. Regards, Roger -- Roger Leigh Printing on GNU/Linux? http://gimp-print.sourceforge.net/ Debian GNU/Linuxhttp://www.debian.org/ GPG Public Key: 0x25BFB848. Please sign and encrypt your mail. signature.asc Description: Digital signature
Bug#318525: FTBFS: Memory inputs not directly addressable
tags 318525 + confirmed thanks I've confirmed the bug still exists with the current GCC 4.0.1 in a sid chroot. The error is identical to that reported by Matt in the original report. Regards, Roger -- Roger Leigh Printing on GNU/Linux? http://gimp-print.sourceforge.net/ Debian GNU/Linuxhttp://www.debian.org/ GPG Public Key: 0x25BFB848. Please sign and encrypt your mail. -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]