libre #smartos

2 Jul 2025

SmartOS -- A Modern, Open Source, Datacenter Operating System | smartos.org | github.com/TritonDataCenter/smartos-live | github.com/TritonDataCenter/triton | Security -- TritonDataCenter.com/security-issues | Channel logs - log.omnios.org/smartos

• 11:24
  jperkin
  pjustice: I did not ;) appears I can reproduce it by running the test suite
• 12:58
  jperkin
  which has now led down various rabbit holes, the current one being tcl rand() not outputting the correct format
• 12:59
  jperkin
  should always return between 0 and 1, e.g.
• 12:59
  jperkin
  % expr rand()
• 12:59
  jperkin
  0.21433575740751612
• 12:59
  jperkin
  but on us it prints
• 12:59
  jperkin
  % expr rand()
• 12:59
  jperkin
  -4.059115371007496e-37
• 12:59
  jperkin
  I can't remember if that's a libm/formatting thing or just a tcl bug?
• 13:28
  pjustice
  that's... neat
• 13:32
  pjustice
  Later last night, I found some similar behavior reported in a redis ticket, and added a note there. This morning they say my problem is different, but they have a fix. Problem related to evport. redis/redis #14162
• 13:51
  jperkin
  ok cool, that's definitely where I was seeing a crash
• 13:54
  jperkin
  just tried that patch and it's not core dumping any more (though still seeing the tcl-related errors)
• 13:55
  jperkin
  I'll get that pkgsrc patched
• 13:57
  pjustice
  many thanks!
• 15:22
  jperkin
  (interesting looks like the tcl thing is a gcc14 issue)
• 15:24
  pjustice
  The inexorable regression of gcc :)
• 15:46
  jperkin
  heh, well I don't know if this is what's ultimately causing the problem, but diff'ing the gcc13 build, for some reason building with gcc14 makes tcl think we're big endian...
• 15:48
  jperkin
  that's definitely in the "special" category