IRC channel logs


back to list of logs

<Pellescours>damo22: I made progress on building tools of netbsd
<Pellescours>I fixed the root cause that was making code entering in cpp_errno_filename (which segfault)
<Pellescours>the root cause is because read fails and should not. I’m trying to understand and reproduce it in a simple example
<damo22>that's good
<Pellescours>my fix was (just before the read) close and reopen the file. I’m trying to know if it’s because of the O_NONBLOCK in open that read fails
<Pellescours>because this fix is really hacky
<damo22>maybe remove the O_NONBLOCK
***Server sets mode: +nt
***Server sets mode: +nt
<Pellescours>WTF, I found the reason
<Pellescours>it’s because of a ternary check to decide of flag to set or not. If I remove the ternary check it work
<Pellescours>here how to reproduce
<Pellescours>that’s really strange
<Pellescours>Ah, I found:
<Pellescours>no not found
<Pellescours>22 int a = O_RDONLY | O_NOCTTY | (val != NULL) ? O_NONBLOCK : 0;
<Pellescours>23 int b = O_RDONLY | O_NOCTTY | O_NONBLOCK;
<Pellescours>24 int c = O_RDONLY | O_NOCTTY | 0;
<Pellescours>25 int file = open(filename, O_RDONLY | O_NOCTTY | (val != NULL) ? O_NONBLOCK : 0, 0666);
<Pellescours>(gdb) p a
<Pellescours>$1 = 8
<Pellescours>(gdb) p b
<Pellescours>$2 = 9
<Pellescours>(gdb) p
<Pellescours>(gdb) p c
<Pellescours>$3 = 1
<youpi>put parentheses
<youpi>?: is terribly low priority
<youpi>so | goes into the parameter of ?
<youpi>instead of getting combined with the ?: result
<Pellescours>that’s it
<Pellescours>I send a patch to NetBSD to have this fixed
<Pellescours>good night, and thank you youpi
<Pellescours> actually gcc seems to have the same problem (I don’t ko
<Pellescours>I don’t know why problem did not appear before (gcc is built on hurd so it should have trigger this problem)
<Pellescours>I know. It’s because on gcc they don’t have this ternary operation
<dogtooth>I'm interested in hanging around and doing Hurd stuff. I'm still pretty new so I'll just try to chase a bug here or there and ask questions when I get stuck.
<dogtooth>I was going to see about GSoC but I understand the deadline has already passed by.
<dogtooth>The Hurd is very cool, I'm enjoying playing with it in qemu right now
***rekado_ is now known as rekado
<gnu_srs1>damo22: Not much is happening recently. Any ideas on how to make at least the rumpdisk boot startup/shutdown working?
<youpi>things did happen recently
<youpi>see the mails on bug-hurd
<damo22>dogtooth: thats cool, welcome
<damo22>i cant do much work on hurd this week, i am moving house
<damo22>that said, we're getting closer to fixing the stat issue, and that will help shutdown/startup
<Pellescours>now compile blocks with CLOCK_REALTIME not defined :D
<youpi>that's a missing #include or #define
<youpi>we do have that macro
<Pellescours>it’s because the corresponding include is done only if __linux__ is defined
<Pellescours>it’s __gnu__ for hurd?
<youpi>it's __GNU__
<youpi>what include ?
<youpi>that's just completely posix
<youpi>there is no reason to use ifdefs for such thing
<Pellescours>I don’t know why but there is an ifdef
<youpi>what include?
<youpi>really there is no reason to have an ifdef, time.h is in posix since ages
<youpi>possibly that conflicts with definitions above, but a #else would be enough
<Pellescours>that’s NetBSD codebase
<youpi>yes but you can use a #else
<Pellescours>I did the change and it worked
<Pellescours>I built tools successfully
<Pellescours>AND I built Netbsd kernels \o/
<Pellescours>so except the two tools issues (patches are sent) we should no encounter problem for building rump
<dogtooth>hey guys, anyone still around?
<Pellescours>yes dogtooth
<dogtooth>very cool @Pellescours, just trying to get a sense of when this chat is active
<ThinkT510>dogtooth: there is a link in the /topic for channel logs too
<dogtooth>ah, thank you ThinkT510!
***Emulatorman_ is now known as Emulatorman