Groups | Search | Server Info | Login | Register
Groups > muc.lists.netbsd.tech.userlevel > #11755
| From | Valery Ushakov <uwe@stderr.spb.ru> |
|---|---|
| Newsgroups | muc.lists.netbsd.tech.userlevel |
| Subject | libphkmalloc? (Was: CVS commit: src/lib/libbsdmalloc) |
| Date | 2026-04-28 19:33 +0300 |
| Organization | Newsgate at muc.de e.V. |
| Message-ID | <afDhXVxpdLygT8bn@snips.stderr.spb.ru> (permalink) |
| References | <20260428130210.539E3FA26@cvs.NetBSD.org> |
On Tue, Apr 28, 2026 at 13:02:10 +0000, Taylor R Campbell wrote: > Module Name: src > Committed By: riastradh > Date: Tue Apr 28 13:02:10 UTC 2026 > > Modified Files: > src/lib/libbsdmalloc: malloc.c > > Log Message: > libbsdmalloc: Use mutex_init, not mutex_unlock, in child post-fork. [...] I wonder if it's time to retire libbsdmalloc? Now that we have had jemalloc in libc for ages, may be it makes sense to provide libphkmalloc instead? It's a derivative of bsdmalloc that fixed a bunch of problems (see phk's history writeup). It's the old-style sbrk-based malloc that might actually still be useful for older software - jemalloc can be instructed to use sbrk, but it uses TLS and that might conflict with older sofware that was written before TLS and assumes some registers (that TLS ABI now uses) are available to apps (e.g. I had to use it on NetBSD/sparc for Self (selflanguage.org) b/c of this). If we are going to go an extra mile, we migth as well do it with the code that is actually useful :). PS: Why bsdmalloc.3 is not insalled, btw? -uwe -- Posted automagically by a mail2news gateway at muc.de e.V. Please direct questions, flames, donations, etc. to news-admin@muc.de
Back to muc.lists.netbsd.tech.userlevel | Previous | Next — Next in thread | Find similar
libphkmalloc? (Was: CVS commit: src/lib/libbsdmalloc) Valery Ushakov <uwe@stderr.spb.ru> - 2026-04-28 19:33 +0300 Re: libphkmalloc? (Was: CVS commit: src/lib/libbsdmalloc) Jason Thorpe <thorpej@me.com> - 2026-04-28 10:13 -0700
csiph-web