IRC channel logs

2023-03-26.log

back to list of logs

<gforce_de1977>aggihmmm, it seems you are going a totally different path. and website or gitrepro which is showing that in more detail or at least outlines it with a bigger picture? or still collecting ideas?
<gforce_de1977>aggi: hmmm
<gforce_de1977>aggi: for your oom-problem in rock64: activate at least HDD-swap or zram-swap
<aggi>gforce_de1977: i won't
<gforce_de1977>aggi: does it mean: your are NOT going a different path or your are not activating SWAP?
<aggi>you seem to misunderstand, what the whole problem is when git or gcc hit oom-killer, with 800MiB RAM available
<gforce_de1977>aggi: what IS the underlying problem? miscompiled code? bugs?
<aggi>the problem are linux,git,GNU/gcc
<aggi>and the oom-kill, is merely a symptom, for what's wrong with them
<gforce_de1977>GCC is mabye not optimal, but at least it supports many arch's - but i understand your disappointment
<avih>well, while i'm not familiar with gcc code, many times there's a tradeoff between resources used at runtime and speed of development/progress. optimizing for space/time when cpu and ram are cheap can be considered by some as without much value (obviously beyond good algorithms). it takes time to make the code more efficient, which some prefer to spend on making the software able to do more things. not advocating this or the other, but it's important to
<avih> understand the tradeoffs of being super optimized, e.g. for low ram usage
<aggi>avih: it's some nasty hack inside gcc makefiles, which triggered the oom-kill (genattrtab), not gcc itself
<aggi>with gcc, it is mainly g++ compiler which freaks out over memory consumption, and since all c++ dependencies were removed, there's no need for this
<avih>"<aggi> the problem are linux,git,GNU/gcc"
<aggi>yes, because they fail bare minimum sanity checks
<avih>in general, yes, some things use more ram than they could, for instance git started as shell scripts. that's typically not optimal. but it worked and was developed very quickly, and took over the world. so obviously not everyone thinks it's terrible.
<Irvise_>Just as a comment. Lua 5.4 does build with TCC/mob.
<Irvise_>So does JimTCL and some smaller Scheme implementations.
<Irvise_>ACTION goes to sleep :)
<avih>:)
<Irvise_>ACTION Irvise has still not forgotten about the Ada-bootstrap problem. But he will probably focus on getting a binary-blob first in Guix to start things out.
<Irvise_>Some people look forward to Ada in Guix because of coreboot.
<muurkha>I didn't know about JimTCL!
<Irvise_>Its 95% of TCL, but no Tk support AFAIK. You can see some of its code "stats" in my comparison table: https://irvise.xyz/Blog/scheme-implementation-comparison.html
<Irvise_>I need to look into mksh.
<Irvise_>Okay, now I go to sleep