log☇︎
3200+ entries in 0.265s
ascii_field: mod6: i kinda assumed you and ben_vulpes would roll the patch sequence docs into releases
mod6: there are a number of different directions these emails go in; patching for v0.5.3.1, pogotronic, gentoo, etc, etc. so I'll try my best to make this apparent when I write up a patch list, etc.
assbot: Logged on 03-07-2015 19:44:53; phf: mod6: well, lets say you have something like mutt open with the current email. it's pretty easy to write a script that takes current email and feeds it to an external script, that splits out patch, verifies it with the provided sig, and then applies it to the codebase (or pushes it into some queue of patches as a case may be). i.e. you read the email, push "p" to do everything for you, and then move on to the n
mod6: basically, instead of me making a whole bunch of different one-off scripts for peoples seperate clients and whatever stack they've got, I just create one script that will patch in & verify all the scripts at once. usually after I've tried and tested them all myself, peronsally.
phf: mod6: well, lets say you have something like mutt open with the current email. it's pretty easy to write a script that takes current email and feeds it to an external script, that splits out patch, verifies it with the provided sig, and then applies it to the codebase (or pushes it into some queue of patches as a case may be). i.e. you read the email, push "p" to do everything for you, and then move on to the next email ☟︎
mod6: <+phf> mod6: no no each email. i thought that was the whole point of ascii's approach, i.e. linux kernel style "read email, think, apply the patch" << im not sure what their process is.
phf: mod6: no no each email. i thought that was the whole point of ascii's approach, i.e. linux kernel style "read email, think, apply the patch"
phf: mod6: i don't necessarily have problems, i have a bitcoind with all the patches up to eatblock (haven't looked at stator yet, but then i'm building with enemy tools). the way i assembled patches is by reading through web archive and saving/verifying each patch as i saw it
mod6: ex: are you trying to patch v0.5.3 to get to v0.5.3.1? (You can just download the release tarball). Are you trying to patch v0.5.3.1-RELEASE with alf's patches? (must read emails to figure out the deps, OR you can just build the stator which includes all of alf's patches with exception of dumpblock & eatblock), those must be patched post extraction of the stator tarball -- of which im actually testing now.
mod6: I will work on a patch list and maybe a script later this month. It is a bit hard to follow. ☟︎
phf: mod6: it would be nice to have an mbox or maildir tgz for the mailing list with all the attachments still included. took me couple of hours to reconstruct and verify patch list, while could've been done in matter of minutes with procmail and mimetools
assbot: Buildroot (busybox) - [PATCH 0/5] Support for musl based external toolchains ... ( http://bit.ly/1dADYyn )
asciilifeform: http://buildroot-busybox.2317881.n4.nabble.com/PATCH-0-5-Support-for-musl-based-external-toolchains-td52730.html << buildroot apparently supports it
ronaz: we are also shipping first patch of Denarium coins in a weeks time
assbot: dpaste: 2DDHFF4: jautenim_openssl_docs_snip.patch.sig ... ( http://bit.ly/1f1Ifwl )
assbot: dpaste: 2XV9P7V: jautenim_openssl_docs_snip.patch ... ( http://bit.ly/1f1IebW )
assbot: Logged on 30-06-2015 17:59:00; mod6: ah, i don't think you're in assbot's L2. Sorry if I wasted your time. Just send the patch to me: modsix@gmail.com
decimation: these were made with your thermonuke patch
mod6: I did it like that because the list of patch names is far too long. makes it look kinda ridiculous in the header
shinohai: @ mod6 so no more manual patch applications?
mod6: <+mod6> plz clearsign your email, and attach a detached signature of the patch file << along with the patch itself i might add!
mod6: plz clearsign your email, and attach a detached signature of the patch file
mod6: ah, i don't think you're in assbot's L2. Sorry if I wasted your time. Just send the patch to me: modsix@gmail.com ☟︎
ben_vulpes: how about that eatblock patch?
mod6: <+Jautenim> shall i submit a patch to the mailing list? it's quite a simple fix << If you're in the WoT (looks like you are) feel free to submit a patch as you like.
mod6: Jautenim: sorry, I had it kinda mixed up. The patch link I posted is to resolve the issue where the libs and headers don't get copied over.
Jautenim: but this patch don't seem to avoid the offending makefile directive, does it?
mod6: that patch is here: http://thebitcoin.foundation/ml/btc-dev/2015-April/000082.html << this patch does a few other things too, might be worth a read through though to see how it aligns with what you had already done to your own local auto.sh.
mod6: nubs submitted a patch for that -- it'll be fixed in the next milestone for sure. thanks for the heads up Jautenim.
Jautenim: shall i submit a patch to the mailing list? it's quite a simple fix
decimation: yeah, patch is better, agreed
asciilifeform: decimation: i just now wrote the patch, l0l
asciilifeform: (easiest method is to actually understand the changes, as one ought to in any case, and re-apply them by hand, then produce new patch)
asciilifeform: http://log.bitcoin-assets.com/?date=30-06-2015#1181387 << my igprof patch for therealbitcoin is almost certainly b0rked by the 'dumpblock' addition. but fixing it ought to be trivial - good project for any of the less-experienced folks here ☝︎
asciilifeform: BingoBoingo: tried the block dumper patch yet ?
decimation: asciilifeform: yeah that's the kind of patch I was implying
asciilifeform: mircea_popescu: see log. new patch
ben_vulpes: line number of patch
ben_vulpes wonders at actual utility of a patch shaped like this
jurov: asciilifeform: ^ we're back to using our own patch hashes somehow
ascii_field: notice that you don't even need the patch file for this
mod6: so after review you just expect or perhaps exepect to just see a .sig file attached alone without the patch itself?
jurov: and user is unwilling to work with patch ID the system must somehow heuristically
mod6: anyway, i like the idea of a checksum for the patches. if we can somehow hvae that without it being in the name of the patch, that would be neat.
ascii_field: but for those discussions to make sense, the patch had to be out there
mod6: i think the nomenclature is pretty straightforward. easy to use. you specify which base your coming off of, the patch name, and a number for any that are required to come before it.
mod6: this is supposed to outline how to name and number the patch: http://therealbitcoin.org/ml/btc-dev/2014-December/000022.html
ascii_field: where is this 'turd.patch' ?
jurov: much better that turd.patch
ascii_field: how is asciilifeform_ver_now_5_4_and_irc_is_gone_and_now_must_give_ip_ebed1af0253ef629bbef4bf2b2d1a94742a81f0e.patch an improvement over asciilifeform_ver_now_5_4_and_irc_is_gone_and_now_must_give_ip.patch ?
jurov: fellows, we have with mod6 noticed disconnect between patch IDs in mailing list and in release notes
ascii_field: mod6: please post the net.cpp immediately prior to the failed patch
ascii_field: shinohai: when you apply that patch, please be sure to read the instructions
shinohai: I still haven't done the irc patch. I need to stop being lazy
mod6: i'll take it down, and add the irc patch
ascii_field: thinking that mod6 will roll them up into a 'static fix' patch
mod6: (11:31) <+mod6> asciilifeform: I just did the following: extracted v0.5.3.1-RELEASE applied the following patches successfully { dnsseed_snipsnip, kills-integer-retardation, nubs-gentoo-sanity, orphange-thermonuke, orphange-tx-amputation, dns-thermonyukyoolar } but when I added the patch for IRC demo, got the following error:
mod6: ascii_field: oh, i did have an issue patching in that last IRC demo patch.... im gonna try it again, but it complained about net.cpp:ThreadGetMyExternalIP
mod6: asciilifeform: I just did the following: extracted v0.5.3.1-RELEASE applied the following patches successfully { dnsseed_snipsnip, kills-integer-retardation, nubs-gentoo-sanity, orphange-thermonuke, orphange-tx-amputation, dns-thermonyukyoolar } but when I added the patch for IRC demo, got the following error:
mod6: But anyone is welcome to patch on their own.
mod6: i've tried like 3 different versions of gcc, with and without my special gcc patch shoehorned in, 2 different versions of uclibc. not quite sure what to think yet.
mod6: if we do anything else other than 5.4, that'd probably be good cause otherwise we'll hvae to patch your patch. :]
asciilifeform: i'll sign an identical patch sans the bump if anyone needs this.
mod6: yeah, as far as the version number, in hindsight; going forward, we should probably keep all version bumps in it's own patch incase something changes. i.e. your latest IRC Full Demolition Patch.
decimation: ah I see, was looking at earlier patch
decimation: phf: I made a patch to note from which node each block comes
asciilifeform: phf: somewhere we have an anti-debug.log-rollover patch
asciilifeform: phf: if you generated this with a custom patch for the beast itself, please share this
mod6: So anyway, we'll have to remember this when we revist the Checkpoints issue. My patch will be rejected then, and we'll come up with something that gets rid of line 939 as well as pulling the checkpoints themselves out of checkpoints.cpp and into a separate config file.
ascii_field: and was forgotten when bdb locks patch discovered
mod6: ascii_field: re: checkpoints: the patch worked fine to remove checkpoints, but it was tabled for the time being. the main discussion around this was that it could be helpful to hvae checkpoints in there to prevent spamming of blocks from timestamps before a checkpoint. ☟︎
ascii_field: ben_vulpes, mod6, mircea_popescu, et al: can anyone recall why http://therealbitcoin.org/ml/btc-dev/attachments/20141218/bitcoin-v0_5_3-rm_checkpoints_41327b9a962e6d27869f4d361d742ab5c7061ede.5.patch didn't make it in ? ☟︎☟︎
mod6: if I add a simple patch file for the -fPIC to /etc/portage/patches/sys-lib/<patchfile> and then emerge uclibc, then it builds, but the outcome is the same of the bitcoind compilation
mod6: well, so here's the deal. I can't get it to compile "by hand". meaning, that if i extract the 2 bzip'd files for uclibc: uClibc-0.9.33.2.tar.bz2 & uClibc-0.9.33.2-patches.tar.bz2 and then patch the former with the latter with something like this: for i in `ls ../patch/*.patch | sort` ; do patch -p1 < ../patch/$i ; done
trinque: mod6: could patch the Makefile using that patches mechanism
mod6: guess like we're gonna have to get this patch to work
mod6: If this dones't help (i doubt it will), I'll need to make another AMI with a recent stage3 and try to shoehorn in this gcc patch I made.
mod6: if that doens't work, then truly will have to get everything in sync and shoe-horn our own patch while all the stage3 stuff is up to speed.
mod6: Anyway, trinque wasn't even able to apply my patch successfully on his side.
mod6: So this is interesting: I was able to use my gcc patch successfully against my AWS instance which has an AMI that utilizes a stage3-uclibc-hardened from 20150510. And meanwhile, trinque was helping me test to ensure I'm not doing something totally retarded, and on his side he was using the Physical hardware steps where it says to pull the latest (which should be 20150610) stage3-uclibc-hardened.
ben_vulpes: <asciilifeform> http://therealbitcoin.org/ml/btc-dev/2015-June/000101.html << well since this thermonuked node refuses to attach to any others, i may as well roll this patch in
mod6: as soon as I can, i'll be sure to patch everything in the order specified and kick off a full regression cycle.
mod6: ok weird. same exact error as last time after complete nuke. tomorrow i'll nuke again and see if gcc will compile without the patch.
mod6: i'm gonna nuke that machine again and try to compile it again with the patch. see if it works. if not, i'll just see if it'll even compile gcc itself.
mod6: here's the gcc 4.8.4 patch in question (gcc needs to be patched with Gentoo/uclibc/PIE patches beforehand): http://dpaste.com/0QHZ59X.txt
mod6: eh, guess i spoke too soon: my patch comiled in nicely... and much much later in the entire compile process gcc hangs on libcpp: http://dpaste.com/3XZH8PZ.txt
mod6: oh, and I keep hitting an additional problem aside from what the "gentoo sanitiy" resolves. i've been holding off submitting a patch for it until I can get this thing actually compiled and working.
mod6: oh hey, i think my patch built ok...
mod6: The patch we got from that mailing list was from a different version for sure than 4.8.4 -- so i've had to hack it in by hand. gentoo is now applying the patch properly. and it's compiling as we speak. there is one caviat that I need to get over before I can test it though.
ben_vulpes: ;;later tell asciilifeform can't find my fat-fingering, but my node running your patch *does* connect to lfnet. it does not, however, obtain a connection to another node thereby.
BingoBoingo: mircea_popescu: The Todd replace by fee patch is probably derp to hell, but it leads to some of the best Hearnia trolling outside this channel https://twitter.com/petertoddbtc/status/612780261109960704
mod6: patch applies manually just fine, seems to choak when used by emerge becuase of directory levels or something. hopefully can overcome this...
mod6: ok, i've got a patch i created myself (post application of patches: gentoo/uclibc/PIE) for gcc... just trying to get it to work with emerge.
mod6: lol, this is uglier than I had thought. I patched the gcc-4.8.4 source with all of the patches from /their/ bundle v1.5 (http://dpaste.com/1NRJTT7.txt), and now i'm in there apply patches by hand, and only 2 out of ~10 files to patch actually exist here (the rest are for pr# files (Problem Report?)). And when I can apply a patch, the surrounding code is slightly different.
mod6: so trinque & I have discovered that the patch provided by the email mentioned ( https://gcc.gnu.org/ml/gcc-patches/2015-02/msg00410.html ) doesn't patch cleanly at all. Even with some McGuyvering of the patch to ensure the paths are correct etc, there are still a number of files not found. It might be plausible to write our own custom patch for 4.8.4 to resolve the issue.
mod6: So, I think I'm gonna stay the course on trying to patch 4.8.4... if we get into a giant hassle with it, we'll cut bait for the time being and try to build something like 3.7 and try that.
mod6: if not, we might have to McGuyver our own patch.
mod6: decimation: https://gcc.gnu.org/ml/gcc-patches/2015-02/msg00410.html < gcc patch that maybe fixed the issue << i just really hope this applies cleanly, and "works".
mod6: I, with trinque's help, need to patch gcc 4.8.4 with gentoo using /etc/portage/patches via ebuild flag(?). If that works, then I can test that the R.I. will link properly. If that works, maybe we finally have static apple pie.
assbot: Logged on 11-06-2015 03:34:37; decimation: https://gcc.gnu.org/ml/gcc-patches/2015-02/msg00410.html < gcc patch that maybe fixed the issue