IRC channel logs
2023-02-08.log
back to list of logs
<oriansj>Christoph[m]: looking at the uFork architecture, it looks like someone rediscovered the TI-990 architecture <oriansj>but then again, it isn't surprising given its most common form was the TI-99/4A where they didn't even have an editor, assembler, or hardware technical information available... <oriansj>is perl guessing the shell based on running processes? <stikonas>I was more thinking of accidentally reading the wrong file <oriansj>and getting only the start of the line? <stikonas>oriansj: start of the line is probably make <stikonas>it only prints command that is being executed <stikonas>e.g. if that line somehow ends in shebang <stikonas>anyway, hard to tell without anybody actually reproducing that state and examining it <stikonas>there enough information to tell what exactly might have happened <stikonas>oriansj: so now that we have a changelog, maybe once you have enough spare time it's worth tagging new releases? <stikonas>at least I haven't seen any serious issues recently <stikonas>(there are some issues in AArch64 but they are not new regressions) <stikonas>and x86 is extensively tested in live-bootstrap anyway <stikonas>fossy: maybe try adding "-s" to LDFLAGS to reduce size? <fossy>stikonas[m]: hmm let me give that a try <fossy>stikonas[m]: hgue difference if we strip <fossy>hmm, wonder if we should consider adding a stripping hook to the end of src_install? <fossy>rickmasters: what's the cause? its very strange <rickmasters>The help2man.PL file is extracted after help2man.1 from the tar file and <rickmasters>if the seconds of the current time changes between the two, then make will trigger a rule which uses help2man <rickmasters>The solution is to add "touch help2man.1" before running make in help2man-1.36.4.sh <rickmasters>I spent quite a while debugging that problem a few weeks ago. <rickmasters>I was behind on reporting it. It was on my to-do list. <rickmasters>It happened during Fiwix testing where it occurred much more frequently because I/O is a little slower. <rickmasters>But recently I ran into it during a normal chroot build (without Fiwix). <rickmasters>I ended up adding an "ls --full-time" before running make and noticed it only happens when the files in question have different seconds in their mod time <rickmasters>Another solution would be to modify the Makefile.in to fix the shebang at the end of the $(target) rule. <rickmasters>Then, if the man page $(target).1 needs to be regenerated, then help2man will have a fixed shebang. <rickmasters>Anyway, I'm signing off. I'll be around tomorrow and can provide more info if needed. <stikonas[m]>fossy, you can try, just (of course conditionally on strip being installed) <stikonas[m]>src_install would need automatic detection of files to strip <rickmasters>fossy, stikonas: I provided a more detailed explanation of the help2man problem and a direction on how to fix it properly: <rickmasters>You're welcome. I don't know enough about how perl was built to fix it properly, but the workaround has been solid for me. <stikonas>rickmasters: I can try a proper fix, it might be just adding -Dstartperl='#!/my/shebang' to makefile (to CPPFLAGS) would fix it and remove the need for shebang fix <rickmasters>Yes it's worth a try. I'm assuming you mean: -Dstartperl='#!/usr/bin/perl" -Dperlpath='/usr/bin/perl' <rickmasters>stikonas: fixing the quotes: -Dstartperl='#!/usr/bin/perl' -Dperlpath='/usr/bin/perl' <stikonas>should be easy to test, first remove the workaround, build should fail <rickmasters>stikonas: I took the liberty of testing a makefile fix since I finally found the perl makefile <rickmasters>stikonas: it didn't work so I dug deeper and finally found the right solution <rickmasters>stikonas: the variables need to be defined in sysa/perl-5.6.2/files/config.sh <stikonas>you already did most of the work, so I would hope it's not too much extra work to finalize it <rickmasters>stikonas: I don't mind. There will be at least a couple checksum changes. <stikonas>but you can get most of them automatically with --update-checksums <rickmasters>stikonas: yeah. Hopefully just those two. I also tested triggering the man page rule (with touch help2man.PL) and that worked. <stikonas>that will ignore checksum checking and then at the end you can just run sha256sum in /usr/src/repo <rickmasters>My test included removing the shebang fix and all is well so far.