Re: Help with -dbg packages for a library
On Fri, Aug 31, 2007 at 10:57:55AM -0400, Justin Pryzby wrote: On Thu, Aug 30, 2007 at 09:29:45AM +0530, Kumar Appaiah wrote: Dear Debian Mentors, I have a specific question with regard to -dbg packages for libraries. My understanding of generating -dbg libraries is like this: 1. We build the package with CFLAGS or CXXFLAGS = -g -O2 (for optimization). 2. We call dh_strip while exluding the dbg package, to ensure that debugging sumbols are present there. I think the suggestion is for all libraries to use dh_strip --dbg-package or -k. BTW this is in bug #420540. Justin -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]
Re: Help with -dbg packages for a library
On Fri, Aug 31, 2007 at 10:57:55AM -0400, Justin Pryzby wrote: So I think there should be no performance difference between running with the libraries compiled without -g, compiled with -g, compiled with -g and stripped, and compiled with -g and debug symbols/sections moved to a separate file. Thanks for the detailed explanation. It would be neat if you could compare the ELF files using binutils tools. I didn't do that, but ran some tests myself (it's a mathematical library), and concluded that there's little to chose from between -g + stripping and no debug+no stripping. Thanks! Kumar -- Kumar Appaiah, 458, Jamuna Hostel, Indian Institute of Technology Madras, Chennai - 600 036 signature.asc Description: Digital signature
Re: Help with -dbg packages for a library
On Thu, Aug 30, 2007 at 09:29:45AM +0530, Kumar Appaiah wrote: Dear Debian Mentors, I have a specific question with regard to -dbg packages for libraries. My understanding of generating -dbg libraries is like this: 1. We build the package with CFLAGS or CXXFLAGS = -g -O2 (for optimization). 2. We call dh_strip while exluding the dbg package, to ensure that debugging sumbols are present there. I think the suggestion is for all libraries to use dh_strip --dbg-package or -k. Now, my situation is that upstream generates special pkg_debug packages by sending ./configure --enable-debug. While I achieve the desired result with the CFLAGS mentioned above, upstream fears that generating the library with debugging symbols first and then stripping them may result in a slightly reduced performance (it's a numerical computation library). While I am going to run some tests myself to verify this, I just wanted to ask the mentors here about their knowledge of this issue. My *understanding* is that the debug information is in a separate ELF section in the executable. See the strip manpage for details, but you can move the debug sections to a separate file, and distribute such files as separate Debian packages. So people who use the library as just a dependency of some other package (and don't have problems so don't need to debug it) just get the lib, and people who develop with it get lib-dev, and also lib-dbg. So I think there should be no performance difference between running with the libraries compiled without -g, compiled with -g, compiled with -g and stripped, and compiled with -g and debug symbols/sections moved to a separate file. It would be neat if you could compare the ELF files using binutils tools. Justin -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]
Help with -dbg packages for a library
Dear Debian Mentors, I have a specific question with regard to -dbg packages for libraries. My understanding of generating -dbg libraries is like this: 1. We build the package with CFLAGS or CXXFLAGS = -g -O2 (for optimization). 2. We call dh_strip while exluding the dbg package, to ensure that debugging sumbols are present there. Now, my situation is that upstream generates special pkg_debug packages by sending ./configure --enable-debug. While I achieve the desired result with the CFLAGS mentioned above, upstream fears that generating the library with debugging symbols first and then stripping them may result in a slightly reduced performance (it's a numerical computation library). While I am going to run some tests myself to verify this, I just wanted to ask the mentors here about their knowledge of this issue. Thanks! Kumar -- Kumar Appaiah, 458, Jamuna Hostel, Indian Institute of Technology Madras, Chennai - 600 036 signature.asc Description: Digital signature