Re: Strange problem: Builds in container, not on machine
On Thu, Mar 23, 2023 at 4:30 PM Ron Olson wrote: > > Are you referring to the directories under rpmbuild? If so, I delete and > recreate the directory and its structure every single time as part of my > build shell script. If you build with "rpmbuild -bb", then yes, I meant those directories, plus wherever your sources are in case they are on disk uncompressed. I don't think anything owned by "root" should be getting messed up by running builds (I hope you're not running them as root!), but you could try to verify that things aren't broken by running something like "sudo rpm -Va" and checking whether any GCC headers are messed up (i.e. MD5 checksum mismatches). Fabio ___ devel mailing list -- devel@lists.fedoraproject.org To unsubscribe send an email to devel-le...@lists.fedoraproject.org Fedora Code of Conduct: https://docs.fedoraproject.org/en-US/project/code-of-conduct/ List Guidelines: https://fedoraproject.org/wiki/Mailing_list_guidelines List Archives: https://lists.fedoraproject.org/archives/list/devel@lists.fedoraproject.org Do not reply to spam, report it: https://pagure.io/fedora-infrastructure/new_issue
Re: Strange problem: Builds in container, not on machine
Are you referring to the directories under rpmbuild? If so, I delete and recreate the directory and its structure every single time as part of my build shell script. On 23 Mar 2023, at 10:17, Fabio Valentini wrote: On Wed, Mar 22, 2023 at 4:16 PM Ron Olson wrote: The irony is that I figured I had just screwed something up on my system and was going to reinstall when this GH report came in. I’m going to try to reproduce with installing F36, building Swift, then upgrading to F37 and see if that has any effect. I've seen similar problems in the distant past when I needed to maintain C / C++ code for some of my packages ... They were almost always caused by "dirty" (or not completely cleaned) build directories, or were caused by stale items in ccache. Neither "diry build directory" nor ccache affect mock builds (unless you explicitly enable the ccache plugin, which is turned off by default), they seem like likely culprits :) Fabio ___ devel mailing list -- devel@lists.fedoraproject.org To unsubscribe send an email to devel-le...@lists.fedoraproject.org Fedora Code of Conduct: https://docs.fedoraproject.org/en-US/project/code-of-conduct/ List Guidelines: https://fedoraproject.org/wiki/Mailing_list_guidelines List Archives: https://lists.fedoraproject.org/archives/list/devel@lists.fedoraproject.org Do not reply to spam, report it: https://pagure.io/fedora-infrastructure/new_issue___ devel mailing list -- devel@lists.fedoraproject.org To unsubscribe send an email to devel-le...@lists.fedoraproject.org Fedora Code of Conduct: https://docs.fedoraproject.org/en-US/project/code-of-conduct/ List Guidelines: https://fedoraproject.org/wiki/Mailing_list_guidelines List Archives: https://lists.fedoraproject.org/archives/list/devel@lists.fedoraproject.org Do not reply to spam, report it: https://pagure.io/fedora-infrastructure/new_issue
Re: Strange problem: Builds in container, not on machine
On Wed, Mar 22, 2023 at 4:16 PM Ron Olson wrote: > > The irony is that I figured I had just screwed something up on my system and > was going to reinstall when this GH report came in. > > I’m going to try to reproduce with installing F36, building Swift, then > upgrading to F37 and see if that has any effect. I've seen similar problems in the distant past when I needed to maintain C / C++ code for some of my packages ... They were almost always caused by "dirty" (or not completely cleaned) build directories, or were caused by stale items in ccache. Neither "diry build directory" nor ccache affect mock builds (unless you explicitly enable the ccache plugin, which is turned off by default), they seem like likely culprits :) Fabio ___ devel mailing list -- devel@lists.fedoraproject.org To unsubscribe send an email to devel-le...@lists.fedoraproject.org Fedora Code of Conduct: https://docs.fedoraproject.org/en-US/project/code-of-conduct/ List Guidelines: https://fedoraproject.org/wiki/Mailing_list_guidelines List Archives: https://lists.fedoraproject.org/archives/list/devel@lists.fedoraproject.org Do not reply to spam, report it: https://pagure.io/fedora-infrastructure/new_issue
Re: Strange problem: Builds in container, not on machine
The irony is that I figured I had just screwed something up on my system and was going to reinstall when this GH report came in. I’m going to try to reproduce with installing F36, building Swift, then upgrading to F37 and see if that has any effect. On 21 Mar 2023, at 5:17, Jonathan Wakely wrote: On Mon, 20 Mar 2023 at 14:10, Ron Olson wrote: Hey all- I got this issue in my GH account I use for building Swift for Fedora: https://github.com/tachoknight/swift-lang-packaging-fedora/issues/2. The TL;DR is that the person was trying to build Swift on Rawhide which he moved to from an older version of Fedora. This tracks with something I discovered as well: I have a current Fedora VM (not Rawhide) that I’ve been upgrading from version to version for several years now, and suddenly it cannot build Swift due to a series of bizarre error messages like: … /usr/lib/gcc/x86_64-redhat-linux/12/../../../../include/c++/12/set:60:10: note: in file included from /usr/lib/gcc/x86_64-redhat-linux/12/../../../../include/c++/12/set:60: #include ^ /usr/lib/gcc/x86_64-redhat-linux/12/../../../../include/c++/12/bits/stl_tree.h:2008:5: error: missing '#include "gcc/x86_64-redhat-linux/12/../../../../include/c++/12/bits/stl_pair.h"'; 'pair' must be declared before it is used pairHowever, using mock or a container, it builds fine, Rawhide, F37, whatever. I’m guessing there’s a setting or environment variable or … ? No, I don't think so. There's no environment variable that is needed to stop the C++ standard library headers being completely broken. The errors above and in the GH issue make no sense to me, maybe it's a bug in Clang's modules that allow a header to be reincluded, but I don't know why it only shows up on your VM and the reporter's system. ___ devel mailing list -- devel@lists.fedoraproject.org To unsubscribe send an email to devel-le...@lists.fedoraproject.org Fedora Code of Conduct: https://docs.fedoraproject.org/en-US/project/code-of-conduct/ List Guidelines: https://fedoraproject.org/wiki/Mailing_list_guidelines List Archives: https://lists.fedoraproject.org/archives/list/devel@lists.fedoraproject.org Do not reply to spam, report it: https://pagure.io/fedora-infrastructure/new_issue___ devel mailing list -- devel@lists.fedoraproject.org To unsubscribe send an email to devel-le...@lists.fedoraproject.org Fedora Code of Conduct: https://docs.fedoraproject.org/en-US/project/code-of-conduct/ List Guidelines: https://fedoraproject.org/wiki/Mailing_list_guidelines List Archives: https://lists.fedoraproject.org/archives/list/devel@lists.fedoraproject.org Do not reply to spam, report it: https://pagure.io/fedora-infrastructure/new_issue
Re: Strange problem: Builds in container, not on machine
Oh sorry, this was copied from my Fedora 37 machine, with GCC 12. Same error in the same place though. :\ On 20 Mar 2023, at 15:58, Florian Weimer wrote: * Ron Olson: Hey all- I got this issue in my GH account I use for building Swift for Fedora: https://github.com/tachoknight/swift-lang-packaging-fedora/issues/2. The TL;DR is that the person was trying to build Swift on Rawhide which he moved to from an older version of Fedora. This tracks with something I discovered as well: I have a current Fedora VM (not Rawhide) that I’ve been upgrading from version to version for several years now, and suddenly it cannot build Swift due to a series of bizarre error messages like: … /usr/lib/gcc/x86_64-redhat-linux/12/../../../../include/c++/12/set:60:10: note: in file included from /usr/lib/gcc/x86_64-redhat-linux/12/../../../../include/c++/12/set:60: #include ^ /usr/lib/gcc/x86_64-redhat-linux/12/../../../../include/c++/12/bits/stl_tree.h:2008:5: error: missing '#include "gcc/x86_64-redhat-linux/12/../../../../include/c++/12/bits/stl_pair.h"'; 'pair' must be declared before it is used pair/usr/lib/gcc/x86_64-redhat-linux/12/../../../../include/c++/12/bits/stl_pair.h:185:12: note: declaration here is not visible struct pair … Rawhide has GCC 13, so the paths look wrong. Try checking package versions, and run “rpm -Va” to check for corruption. For the error with the /13/ in the path, is it possible that the Clang bundled with Swift cannot handle current libstdc++ headers yet? Thanks, Florian___ devel mailing list -- devel@lists.fedoraproject.org To unsubscribe send an email to devel-le...@lists.fedoraproject.org Fedora Code of Conduct: https://docs.fedoraproject.org/en-US/project/code-of-conduct/ List Guidelines: https://fedoraproject.org/wiki/Mailing_list_guidelines List Archives: https://lists.fedoraproject.org/archives/list/devel@lists.fedoraproject.org Do not reply to spam, report it: https://pagure.io/fedora-infrastructure/new_issue
Re: Strange problem: Builds in container, not on machine
On Mon, 20 Mar 2023 at 14:10, Ron Olson wrote: > Hey all- > > I got this issue in my GH account I use for building Swift for Fedora: > https://github.com/tachoknight/swift-lang-packaging-fedora/issues/2. The > TL;DR is that the person was trying to build Swift on Rawhide which he > moved to from an older version of Fedora. This tracks with something I > discovered as well: I have a current Fedora VM (not Rawhide) that I’ve been > upgrading from version to version for several years now, and suddenly it > cannot build Swift due to a series of bizarre error messages like: > > … > /usr/lib/gcc/x86_64-redhat-linux/12/../../../../include/c++/12/set:60:10: > note: in file included from > /usr/lib/gcc/x86_64-redhat-linux/12/../../../../include/c++/12/set:60: > #include > ^ > /usr/lib/gcc/x86_64-redhat-linux/12/../../../../include/c++/12/bits/stl_tree.h:2008:5: > error: missing '#include > "gcc/x86_64-redhat-linux/12/../../../../include/c++/12/bits/stl_pair.h"'; > 'pair' must be declared before it is used > pair ^ > /usr/lib/gcc/x86_64-redhat-linux/12/../../../../include/c++/12/bits/stl_pair.h:185:12: > note: declaration here is not visible > struct pair > … > > However, using mock or a container, it builds fine, Rawhide, F37, whatever. > > I’m guessing there’s a setting or environment variable or … ? > No, I don't think so. There's no environment variable that is needed to stop the C++ standard library headers being completely broken. The errors above and in the GH issue make no sense to me, maybe it's a bug in Clang's modules that allow a header to be reincluded, but I don't know why it only shows up on your VM and the reporter's system. ___ devel mailing list -- devel@lists.fedoraproject.org To unsubscribe send an email to devel-le...@lists.fedoraproject.org Fedora Code of Conduct: https://docs.fedoraproject.org/en-US/project/code-of-conduct/ List Guidelines: https://fedoraproject.org/wiki/Mailing_list_guidelines List Archives: https://lists.fedoraproject.org/archives/list/devel@lists.fedoraproject.org Do not reply to spam, report it: https://pagure.io/fedora-infrastructure/new_issue
Re: Strange problem: Builds in container, not on machine
* Ron Olson: > Hey all- > > I got this issue in my GH account I use for building Swift for Fedora: > https://github.com/tachoknight/swift-lang-packaging-fedora/issues/2. The > TL;DR is that the person was trying to build Swift on Rawhide which he > moved to from an older version of Fedora. This tracks with something I > discovered as well: I have a current Fedora VM (not Rawhide) that I’ve > been upgrading from version to version for several years now, and > suddenly it cannot build Swift due to a series of bizarre error > messages like: > > … > /usr/lib/gcc/x86_64-redhat-linux/12/../../../../include/c++/12/set:60:10: > note: in file included from > /usr/lib/gcc/x86_64-redhat-linux/12/../../../../include/c++/12/set:60: > #include > ^ > /usr/lib/gcc/x86_64-redhat-linux/12/../../../../include/c++/12/bits/stl_tree.h:2008:5: > error: missing > '#include > "gcc/x86_64-redhat-linux/12/../../../../include/c++/12/bits/stl_pair.h"'; > 'pair' must be > declared before it is used > pair ^ > /usr/lib/gcc/x86_64-redhat-linux/12/../../../../include/c++/12/bits/stl_pair.h:185:12: > note: declaration > here is not visible > struct pair > … Rawhide has GCC 13, so the paths look wrong. Try checking package versions, and run “rpm -Va” to check for corruption. For the error with the /13/ in the path, is it possible that the Clang bundled with Swift cannot handle current libstdc++ headers yet? Thanks, Florian ___ devel mailing list -- devel@lists.fedoraproject.org To unsubscribe send an email to devel-le...@lists.fedoraproject.org Fedora Code of Conduct: https://docs.fedoraproject.org/en-US/project/code-of-conduct/ List Guidelines: https://fedoraproject.org/wiki/Mailing_list_guidelines List Archives: https://lists.fedoraproject.org/archives/list/devel@lists.fedoraproject.org Do not reply to spam, report it: https://pagure.io/fedora-infrastructure/new_issue
Strange problem: Builds in container, not on machine
Hey all- I got this issue in my GH account I use for building Swift for Fedora: https://github.com/tachoknight/swift-lang-packaging-fedora/issues/2. The TL;DR is that the person was trying to build Swift on Rawhide which he moved to from an older version of Fedora. This tracks with something I discovered as well: I have a current Fedora VM (not Rawhide) that I’ve been upgrading from version to version for several years now, and suddenly it cannot build Swift due to a series of bizarre error messages like: … /usr/lib/gcc/x86_64-redhat-linux/12/../../../../include/c++/12/set:60:10: note: in file included from /usr/lib/gcc/x86_64-redhat-linux/12/../../../../include/c++/12/set:60: #include ^ /usr/lib/gcc/x86_64-redhat-linux/12/../../../../include/c++/12/bits/stl_tree.h:2008:5: error: missing '#include "gcc/x86_64-redhat-linux/12/../../../../include/c++/12/bits/stl_pair.h"'; 'pair' must be declared before it is used pair/usr/lib/gcc/x86_64-redhat-linux/12/../../../../include/c++/12/bits/stl_pair.h:185:12: note: declaration here is not visible struct pair … However, using mock or a container, it builds fine, Rawhide, F37, whatever. I’m guessing there’s a setting or environment variable or … ? that is present on my and his systems that doesn’t exist in a new container and thus I nor he can build it, while a container and mock can (and I’ve confirmed with scratch koji builds multiple times). Might anyone have an idea where to check to see what might be causing this? I can’t pin down exactly when this started to happen except that it has been _relatively_ recently, perhaps just in the past few months. Thanks for any suggestions! Ron___ devel mailing list -- devel@lists.fedoraproject.org To unsubscribe send an email to devel-le...@lists.fedoraproject.org Fedora Code of Conduct: https://docs.fedoraproject.org/en-US/project/code-of-conduct/ List Guidelines: https://fedoraproject.org/wiki/Mailing_list_guidelines List Archives: https://lists.fedoraproject.org/archives/list/devel@lists.fedoraproject.org Do not reply to spam, report it: https://pagure.io/fedora-infrastructure/new_issue