***scs is now known as Guest81063
***scs is now known as Guest13375
***scs is now known as Guest28582
***scs is now known as Guest91961
***scs is now known as Guest53508
<fossy>132K patches later, kexec-tools is done <fossy>only 2k loc patches tho, that's not too bad <Hagfish>assuming it works, what happens next? ***scs is now known as Guest33678
***^_ is now known as V
***deltab_ is now known as deltab
***vup2 is now known as vup
***scs is now known as Guest28120
<pder>stikonas: With perl 5.6.2, I tried replacing the config.h you created with one generated from running Configure. miniperl builds but segfaults doing miniperl configpm lib/Config.pm. It actually always crashes. Did you encounter anything like this? <stikonas>pder: no... maybe you need to take config.sh too? <pder>Replace config.sh with the generated one? <pder>I can try that. How did you arrive at the config.h that you used? <stikonas>well, started with earlier config.h from perl 5.5 <stikonas>and addign stuff to it until it compiled <pder>something is very strange because I can set a breakpoint in main in gdb and single step and it steps through unrelated code in perly.h <stikonas>hmm, it shouldn't be doing too many optiomizations <pder>if I rebuild perl after the 3rd tcc musl rebuild the build completes and miniperl works <pder>Oh and autoconf-2.57 now builds <pder>thats with full config.h- so now I can try to update your original config.h with whatever was missing <stikonas>I tried FORK/VFORK but without musl rebuild... <pder>Could perl-5.6.2 be moved after tcc-musl v3? Or is it needed for the other autoconf versions? <stikonas>that means we need to go with older libtool though <stikonas>pder: if you can try too see what's the newest autoconf that works <stikonas>newer gawk might need quite a bit newer autoconf <stikonas>although, I think new autoconf will also print that gawk error on configure <pder>It's rather tricky- I think things would be simpler if we just had binutils ar earlier <stikonas>it might be hard to build it without autotools <pder>I wonder what perl version is needed for autoconf-2.52 <stikonas>pder: autoconf 2.52 does not use perl yet <stikonas>pder: looks like quite low requirements on perl version <stikonas>although, I'm not sure if we install enough things in older versions, but that's fixable\ <pder>I want to test moving perl-5.6.2 to after tcc musl v3 <stikonas>pder: so I guess try moving build after binutils <stikonas>and then I hope I can solve libtool issue ***scs is now known as Guest72330
<pder>stikonas: I can get build up to autoconf 2.61 by running configure followed by make <pder>But we would probably need to incrementally build each version running autoconf and/or automake until we get to this version <pder>After 2.61 I start to get gawk errors on configure <stikonas>well, not sure if it's the same, but besically I used sed to replace placeholder variables <stikonas>but I don't know if it will work with more complex autoconf <stikonas>pder: alternative might be to patch out 2.53 <pder>I was able to move perl 5.6.2 to after tcc musl v3 <pder>And it appears the hang problem has gone away <pder>But I need to simplify config.h <stikonas>well, patching that line out might be less painful <stikonas>than trying to use sed, etc to install autoconf without autotools <pder>Should we keep every version around just in case for later? <stikonas>I have one sed liner that makes them coinstallable <stikonas>2.61 also needs newer automake than what we have <stikonas>so we still need some kind of stepping up ***scs is now known as Guest10881
***scs is now known as Guest28001
<stikonas>now need to check which one of them does it <pder>Ah, could be- I had started to narrow it down <stikonas>just FCNTL gives tcc: error: undefined symbol 'do_aspawn' ***scs is now known as Guest23264
<stikonas>pder: ok, I'll try to move this perl to after musl v3 then <pder>In your testing, were you building perl before or after musl v3? <stikonas>hopefully it doesn't affect any checksums <stikonas>I can try to first do a run without moving perl <pder>For some reason it did make a difference using the config.h generated by Configure <stikonas>if that works, that would be preferable... <stikonas>anyway, started the run, let's see in 20 minutes <pder>I dont know if you noticed this, but when getting that hang in autom4te, I would hit Ctrl-C and get a bunch of garbage output <stikonas>luckily fcntl is early in the alphabet, so I found it really quickly <stikonas>especially after you told that it's stuck in read <stikonas>so I was just going over the headers that have something to do with it <stikonas>Hagfish: decided to first do rerun with updated perl hash <Hagfish>cool, i look forward to the result :) <stikonas>it's just a matter of whether we need to move perl to after musl v3 or just keep it where it is <Hagfish>yeah, that's a weird constraint to manage <stikonas>that will determine when we can get newer autotools, but it's not a particularly big deal <stikonas>well, before binutils I still need to insert libtool... <stikonas>Hagfish: yes, so will be milldly annoying <stikonas>but it does have some checks for ar and ld <stikonas>which I might be able to replace with tcc -ar and true <Hagfish>i get the feeling that the further through the bootstrap process, the bigger the targets, and the more dependencies they have <stikonas>although, it's more of an issue that I've no idea about libtool or how to use it, so need to read some manuals <stikonas>Hagfish: that's true, but we have more stuff already <Hagfish>right, it's nice if the two effects cancel each other out <stikonas>especially when we don't need to write makefiles, find out the right config, etc.. <stikonas>then autoconf/automake/libtool form the whole GNU autotools build system <stikonas>right now we only have old autoconf and old automake <Hagfish>i don't know if there's a community out there that specialises in libtool-related questions <Hagfish>yeah, i guess something so well established should have decent docs that will shed light on any edge case you hit <Hagfish>yeah, no reason to think it's any harder than the existing steps <stikonas>Hagfish: pder: so perl can be kept where it is <stikonas>there is a warning from perl that might be nice to fix <stikonas>Use of uninitialized value in concatenation (.) or string at /after/lib/perl5/5.6.2/Errno.pm line 11. <stikonas>printed by "$Config{'archname'}-$Config{'osvers'}" eq <stikonas>I guess we can add those variables to config.sh (?) if we want