On Thu, 2023-01-26 at 12:41 +0100, Florian Weimer wrote:

• Adam Williamson:

...

6 (__libc_message.cold+0x5) [0x7fbae3c2560f] Jan 25 13:38:47 fedora /usr/libexec/gdm-x-session[1040]: (EE) 6: /lib64/libc.so.6 (malloc_printerr+0x15) [0x7fbae3c96a05] Jan 25 13:38:47 fedora /usr/libexec/gdm-x-session[1040]: (EE) 7: /lib64/libc.so.6 (_int_free+0x9e5) [0x7fbae3c98de5] Jan 25 13:38:47 fedora /usr/libexec/gdm-x-session[1040]: (EE) 8: /lib64/libc.so.6 (__libc_free+0x7e) [0x7fbae3c9b42e] Jan 25 13:38:47 fedora /usr/libexec/gdm-x-session[1040]: (EE) 9: /usr/lib64/dri/zink_dri.so (__driDriverGetExtensions_zink+0x9e70) [0x7fbad82b8180] Jan 25 13:38:47 fedora /usr/libexec/gdm-x-session[1040]: (EE) 10: /lib64/libgbm.so.1 (gbm_format_get_name+0xe81) [0x7fbae3229361] Jan 25 13:38:47 fedora /usr/libexec/gdm-x-session[1040]: (EE) 11: /lib64/libgbm.so.1 (gbm_format_get_name+0x1018) [0x7fbae32294f8] Jan 25 13:38:47 fedora /usr/libexec/gdm-x-session[1040]: (EE) 12: /lib64/libgbm.so.1 (gbm_format_get_name+0x121a) [0x7fbae32296fa]

I saw this during the latest glibc update, but given that the update wasn't tested in isolation, I waived it through because the update addresses the known sprintf issue.

I'm not aware of anything in glibc this might have caused this. The crashes related to -D_FORTIFY_SOURCE=3 look different (they try to crash before causing heap corruption!), and it's not the sprintf assertion failure in __printf_buffer_as_file_commit, either.

No, it's definitely mesa that causes it - I verified that with manual local testing, updating only mesa causes the bug to start happening, downgrading it makes it go away. This is now being tracked/investigated as https://bugzilla.redhat.com/show_bug.cgi?id=2164667 .