Conversation

Notices

1. Embed this notice
   Haelwenn /элвэн/ :triskell: (lanodan@queer.hacktivis.me)'s status on Wednesday, 08-Jan-2025 09:01:06 JST Haelwenn /элвэн/ :triskell:

   • golly josh darn
   • ✧✦✶✷Catherine✷✶✦✧
   • josch

   @klardotsh @josch @whitequark I find it somewhat readable, sure short variable names but strncmp is doomed to be like that.

   The unsigned char cast is interesting, not sure what it's for and I guess could use a comment but it's not really worrying. The printf functions in musl are more annoying to read and I think dalias has made them more readable recently.

   Meanwhile glibc it's:
   - layers and layers of macros for even trivial functions, making it a pain to debug;
   - dangerous features like ifunc and nsswitch.conf (dynamically loading authentication libraries is a hard no for me);
   - ld.so having all kinds of questionable features making it a source of exploits for setuid binaries;
   - python and perl as build dependencies, making it a *very* long adventure to bootstrap from full-source;
   …

   • Embed this notice
     Haelwenn /элвэн/ :triskell: (lanodan@queer.hacktivis.me)'s status on Wednesday, 08-Jan-2025 09:23:49 JST Haelwenn /элвэн/ :triskell:

     in reply to

     • golly josh darn
     • ✧✦✶✷Catherine✷✶✦✧
     • josch

     @whitequark @josch @klardotsh Well ifunc I can see the appeal of it but given the way it's designed I'd rather not have it at all, it's too much of a hack.

     At least I'd much rather use function pointers (which have some overhead), or somewhat bigger functions, or separating needs with different dynamic libraries.

   • Embed this notice
     ✧✦✶✷Catherine✷✶✦✧ (whitequark@mastodon.social)'s status on Wednesday, 08-Jan-2025 09:23:50 JST ✧✦✶✷Catherine✷✶✦✧

     in reply to

     • golly josh darn
     • josch

     @lanodan @josch @klardotsh (ifunc is pretty useful and does not really have a replacement you can deploy yourself)