| 13 Jul 2023 |
|  sorear joined the room. | 23:48:56 |
| 15 Jul 2023 |
 Qubasa | Mic92: This is pretty cool, do you think it would break stuff if we did this by default in NixOS? https://github.com/wheybags/glibc_version_header | 12:19:01 |
 Mic92 | In reply to @qubasa:gchq.icu
Mic92: This is pretty cool, do you think it would break stuff if we did this by default in NixOS? https://github.com/wheybags/glibc_version_header What's the benefit to do this by default. NixOS adds rpath and therefore introduces glibc dependencies. | 17:59:45 |
| Mic92 | * In reply to @qubasa:gchq.icu
Mic92: This is pretty cool, do you think it would break stuff if we did this by default in NixOS? https://github.com/wheybags/glibc_version_header
What's the benefit to do this by default? NixOS adds rpath and therefore introduces glibc dependencies. | 17:59:57 |
| Mic92 | * In reply to Qubasa
Mic92: This is pretty cool, do you think it would break stuff if we did this by default in NixOS? https://github.com/wheybags/glibc_version_header
In reply to @qubasa:gchq.icu
Mic92: This is pretty cool, do you think it would break stuff if we did this by default in NixOS? https://github.com/wheybags/glibc_version_header
What's the benefit to do this by default? NixOS adds rpath and therefore introduces lib dependencies.
| 18:00:14 |
| Qubasa | It would make flakes inputs truly compatible because right now if you have an input that has a mismatching glibc binary you get really difficult to traceback errors | 18:18:55 |
| Qubasa | * It would make flakes inputs truly compatible because right now if you have an input that has a mismatching glibc binary you get really complicated errors | 18:19:28 |
 K900 (deprecated) | Not really | 18:25:09 |
| K900 (deprecated) | If those things still load two glibcs | 18:25:22 |
| K900 (deprecated) | Which they will | 18:25:25 |
| K900 (deprecated) | It'll still explode, just in a different way | 18:25:32 |
| Qubasa | No it shouldn't, glibc ships every old version of a function within, so with this trick everything would just depend on the oldest glibc, and use the oldest functions. Of course it doesn't have to be the oldest function we could also set a sensible minimum version. | 18:55:41 |
| K900 (deprecated) | It still will because we directly stick glibc into every binary's rpath | 18:56:10 |
| K900 (deprecated) | And if we stop doing that, we'll need some weird global ambient glibc | 18:56:21 |
| K900 (deprecated) | Which we explicitly not want | 18:56:28 |
| K900 (deprecated) | * Which we explicitly do not want | 18:56:31 |
| K900 (deprecated) | It will also break reference tracking and such | 18:57:15 |
| Qubasa | In reply to @k900:conduit.0upti.me
It still will because we directly stick glibc into every binary's rpath We can still stick glibc into every rpath, and we don't need a global glibc. What we would need is add a flag to C_FLAGS that includes this magic header file. That's it. | 18:59:54 |
| K900 (deprecated) | No | 18:59:59 |
| K900 (deprecated) | If we stick glibc into every rpath, two different libraries will load two different glibcs | 19:00:13 |
| K900 (deprecated) | And that will still explode | 19:00:17 |
| Qubasa | why so? If they use the same functions? Because of inner glibc state? | 19:01:13 |
| K900 (deprecated) | Because there's a huge pile of global state in glbic | 19:01:25 |
| K900 (deprecated) | * Because there's a huge pile of global state in glibc | 19:01:27 |
| Qubasa | ahhh | 19:01:29 |
| Qubasa | well that sucks | 19:01:35 |
| 16 Jul 2023 |
|  windchimesofmagic joined the room. | 02:36:19 |
| 20 Jul 2023 |
|  fpletz changed their profile picture. | 03:29:41 |
| 24 Jul 2023 |
|  Tamara (dm me on @tammeyy:winter-cottage.eu please) changed their display name from Tamara to Tamara (she/they). | 15:07:48 |
| K900 (deprecated) changed their profile picture. | 15:45:00 |
