IRC channel logs

2022-05-17.log

back to list of logs

<youpi>gnu_srs: gdal built fine, thanks!
<Shentino>So is a memory object just something that is filled on demand as needed by a server and then mapped by the kernel?
<Shentino>and what about CoW?
<luckyluke>Curiosa: on my 64-bit branch I have working IPC for 32-bit userland, now I'm divided between cleaning up the patch set and addressing the next issue booting from a ramdisk
<luckyluke>Curiosa: one thing that could be enabled by your approach is better versioning of the translator API
<luckyluke>but with the vdso-like think you could ask for a specific translator API version
<luckyluke>currently it seems mig doesn't support versioning, if you change one RPC you need to givi it a different msgid
<DiffieHellman>Very nice.
<luckyluke>but I'm still a bit confused how you'll handle the different address spaces... once the application called in the function exported by the translator, it would still use mach_msg and mig, right?
<Curiosa>luckyluke: yes, mach_msg becomes the way to securely access gates to different address spaces
<Curiosa>At least in the thread migration model
<Curiosa>and.. of course.. signals are pure evil
<Curiosa>imagine hanging to such an hack for that long
***FragByte_ is now known as FragByte
***yjftsjthsd31 is now known as yjftsjthsd3
***panzeroceania_ is now known as panzeroceania
<z3ntu[m]>youpi: I'm trying to build glibc with all debian patches now, but it's failing with undefined reference to `__proc_reauthenticate_complete' and I can't find where this function should be implemented, google or salsa gitlab isn't indexing that
<youpi>z3ntu[m]: IIRC it's part of the patches from Sergey, for which I'm waiting for the copyright papers before committing to the repository
<youpi>it's the 00* patches in the debian repository
<youpi>if you think it's crazy that it's still missing upstream, please tell rms
<youpi>he hasn't answered me anything for a few months now
<z3ntu[m]>youpi: but the patches are already built as part of Debian glibc, no? How does it compile then?
<youpi>these are in the hurd package
<z3ntu[m]>I applied both "series" and "series.hurd-i386" or something to glibc-2.33 tag using quilt
<z3ntu[m]>Ah thanks :) gitlab search doesn't find it for some reason
<youpi>note that the heading __ is adding by mig
<youpi>so it's proc_reauthenticate_complete that you need to look for
<Gooberpatrol_66>it sounds like the way the fsf does copyright assignment should be overhauled
<Gooberpatrol_66>that wait is crazy
<youpi>overhauled or just dropped
<Gooberpatrol_66>i mean i'm cool with it if it just takes a few days...
<youpi>legal stuff is prone not to be a few days :)
<Gooberpatrol_66>yeah i'm not a lawyer so i can't fathom what could be so complex about it
<youpi>employers
<youpi>legal names
<youpi>copyright law variety
<amk>why does the copyright assignment still happen? AFAIK even GCC doesn't do it anymore. it took like a month for my copyright assignment to get processed, its a significant barrier
<z3ntu[m]>youpi: so even with glibc patched so it's basically debian glibc, I'm having weird issue. Currently stuck on some issue that if bash does "cat >conf11538subs.awk <<_ACEOF" then like about 1000 lines of stuff, then the normal "_ACEOF" it just hangs forever. If I decrease the lines that should be written to the file then it works fine
<z3ntu[m]>this makes autotools hang after step "configure: creating ./config.status"
<z3ntu[m]>I was hoping that debian patched glibc would fix this :D
<youpi>I don't remember something like that