Hello,
I know this isn't technically Fedora's problem since binary-only drivers are on the ForbiddenItems, but I have been trying to get the nVidia and ATi proprietary to build for the devel/ branch at Livna and there's a problem with the BuildID.
rpmbuild fails because it can't find the BuildID in various files, but how can a BuildID be added to a binary-only file? An example is ATi's /usr/sbin/atieventd, which is a binary executable and there's no place to run `gcc -shared` on it since it's already compiled.
I think this isn't just a graphic driver specific issue but one for all binary-only drivers. It's a perfect example of why they should be avoided, but it's going to be a big problem is _nobody_ can provide binary drivers for F8 because of missing BuildIDs.
Thanks, Stewart
On 8/19/07, Stewart Adam s.adam@diffingo.com wrote:
Hello,
I know this isn't technically Fedora's problem since binary-only drivers are on the ForbiddenItems, but I have been trying to get the nVidia and ATi proprietary to build for the devel/ branch at Livna and there's a problem with the BuildID.
rpmbuild fails because it can't find the BuildID in various files, but how can a BuildID be added to a binary-only file? An example is ATi's /usr/sbin/atieventd, which is a binary executable and there's no place to run `gcc -shared` on it since it's already compiled.
I think this isn't just a graphic driver specific issue but one for all binary-only drivers. It's a perfect example of why they should be avoided, but it's going to be a big problem is _nobody_ can provide binary drivers for F8 because of missing BuildIDs.
is they a way to just not use a buildid? they are used for debugging and we can't debug binary only stuff anyway.
For such a package I don't know useful things the debuginfo rpm might ever have. Why don't you just disable it entirely?
%define _enable_debug_packages 0
Thanks, Roland
On Sunday 19 August 2007, Roland McGrath wrote:
For such a package I don't know useful things the debuginfo rpm might ever have. Why don't you just disable it entirely?
That's probably an ok approach. But with debuginfo packages disabled, rpmbuild does not strip all the binaries which it does with them enabled, so people may want to strip executables and shared libs explicitly during build if the symbols are not useful.
%define _enable_debug_packages 0
That does not disable debuginfo packages for me (F7 x86_64). This does:
%define debug_package %{nil}
On Sunday 19 August 2007, Roland McGrath wrote:
For such a package I don't know useful things the debuginfo rpm might ever have. Why don't you just disable it entirely?
That's probably an ok approach. But with debuginfo packages disabled, rpmbuild does not strip all the binaries which it does with them enabled, so people may want to strip executables and shared libs explicitly during build if the symbols are not useful.
/usr/lib/rpm/redhat/brp-strip is run when find-debuginfo.sh isn't.
%define _enable_debug_packages 0
That does not disable debuginfo packages for me (F7 x86_64). This does:
%define debug_package %{nil}
Bah. Wonders of rpm spec conditionals. Is there an %undef? Maybe the actual rpm wizards involved in setting up the current macro scheme could say what the preferred method is.
Thanks, Roland
On Sunday 19 August 2007, Roland McGrath wrote:
On Sunday 19 August 2007, Roland McGrath wrote:
For such a package I don't know useful things the debuginfo rpm might ever have. Why don't you just disable it entirely?
That's probably an ok approach. But with debuginfo packages disabled, rpmbuild does not strip all the binaries which it does with them enabled, so people may want to strip executables and shared libs explicitly during build if the symbols are not useful.
/usr/lib/rpm/redhat/brp-strip is run when find-debuginfo.sh isn't.
Yes, but unlike find-debuginfo.sh, that one does not strip shared libraries and the like - there's a "grep -v ' shared object,'" in it which explicitly skips them. I haven't checked if there are other differences eg. in handling of executable bits, write permissions, setuid/setgid bits etc.
%define _enable_debug_packages 0
That does not disable debuginfo packages for me (F7 x86_64). This does:
%define debug_package %{nil}
Bah. Wonders of rpm spec conditionals. Is there an %undef? Maybe the actual rpm wizards involved in setting up the current macro scheme could say what the preferred method is.
I think it's the one I posted.
On Mon, 2007-08-20 at 01:33 +0300, Ville Skyttä wrote:
On Sunday 19 August 2007, Roland McGrath wrote:
On Sunday 19 August 2007, Roland McGrath wrote: %define debug_package %{nil}
Bah. Wonders of rpm spec conditionals. Is there an %undef? Maybe the actual rpm wizards involved in setting up the current macro scheme could say what the preferred method is.
I think it's the one I posted.
Yep, %define debug_package %{nil} is what you want
Jeremy