# 00:26:14 |
y_morin |
quits : Quit: Je suis Charlie. |
# 01:09:51 |
mingwandroid |
quits : Quit: Leaving. |
# 03:59:16 |
alan_o |
quits : Quit: Leaving |
# 07:17:26 |
jA_cOp |
quits : Ping timeout: 265 seconds |
# 08:19:20 |
jA_cOp |
joins #crosstool-ng |
# 10:00:17 |
laeckerv |
joins #crosstool-ng |
# 10:11:46 |
y_morin |
joins #crosstool-ng |
# 12:19:16 |
laeckerv |
quits : Ping timeout: 244 seconds |
# 12:22:34 |
mingwandroid |
joins #crosstool-ng |
# 15:00:38 |
smartin__ |
joins #crosstool-ng |
# 15:03:54 |
smartin_ |
quits : Ping timeout: 264 seconds |
# 15:11:23 |
laeckerv |
joins #crosstool-ng |
# 15:26:34 |
laeckerv |
quits : Ping timeout: 265 seconds |
# 16:12:42 |
laeckerv |
joins #crosstool-ng |
# 16:20:23 |
diorcety |
joins #crosstool-ng |
# 18:37:40 |
mnt_real |
quits : Ping timeout: 244 seconds |
# 18:38:50 |
mnt_real |
joins #crosstool-ng |
# 18:55:04 |
mingwandroid |
y_morin: hi. |
# 18:56:58 |
y_morin |
mingwandroid: Hey! |
# 19:01:03 |
mingwandroid |
y_morin: Happy New Year. |
# 19:01:06 |
mingwandroid |
y_morin: brb. |
# 19:01:36 |
y_morin |
mingwandroid: Best Whishes for 2015, too. |
# 19:05:02 |
mingwandroid |
y_morin: so I got a quick question. |
# 19:06:19 |
mingwandroid |
y_morin: I'm having a look at CT_GetGit, and wanted to allow cset to be passed in as, for example, "ref=HEAD". in there, I use git ls-remote to get the SHA1, but I kind of need to pass that back out to the caller so I know the name of the tarball .. |
# 19:06:54 |
mingwandroid |
what do you recommend for that? the usual pass a var name in and use eval dance? |
# 19:08:39 |
y_morin |
mingwandroid: Sorry, I don;t understand why you would need that. Caller must now the version. If you pass a floating reference, there's no guarantee it can be reproduced over time. |
# 19:09:02 |
y_morin |
mingwandroid: If you want to play wih 'bleding-edge' development stuff, it's better to use a local tree of the sources. |
# 19:09:30 |
y_morin |
mingwandroid: And no, there is currently no way to pass back values to a caller. And using the eval-dance is ugly. |
# 19:09:52 |
y_morin |
mingwandroid: Also, that would be the first time an eval is used to pass info back to the caller. |
# 19:10:02 |
mingwandroid |
y_morin: I am updating mingw-w64 stuff, and they don't release nearly often enough, but they have various branches (master, 3.1.0 that's stable ..) |
# 19:10:22 |
mingwandroid |
y_morin: very similar actually to glibc's svn branches that we already handle in a similar way. |
# 19:10:31 |
y_morin |
mingwandroid: Then, directly use a sha1. |
# 19:11:29 |
mingwandroid |
y_morin: you don't see value in just saying "I want master", then having the CT_GetGit figure out the sha1 and return it to the caller for later extraction? |
# 19:11:33 |
y_morin |
(and I am not too happy that it is possible to use svn HEAD revision to start with... If I had have foresight back in the day, I'd have made that impossible) |
# 19:11:50 |
y_morin |
mingwandroid: No, I say that using 'master' is just not stable over time. |
# 19:12:22 |
y_morin |
mingwandroid: If you use 'master' right now, it will give you a sha1. Then you come back tomorrow, it will give you another sha1. |
# 19:12:30 |
mingwandroid |
y_morin: granted, but I do name the tarball with the full sha1, so reproducability isn't a problem. |
# 19:12:34 |
y_morin |
mingwandroid: Thus: toolchain not reprodicible with the same .config. |
# 19:12:59 |
y_morin |
mingwandroid: No, because tomorrow, you'd still be using 'master' in your toolchain's .config. |
# 19:14:01 |
y_morin |
mingwandroid: It is not how you name the local tarball. It's about reproducing the same output (the toolchain) with the same input (the .config). |
# 19:14:29 |
mingwandroid |
y_morin: I could bake the sha1 out into the generated .config, would that be more acceptable or do you just not want such a thing? |
# 19:14:53 |
y_morin |
mingwandroid: Well, I am no longer the maintainer, so what I want does not matter. It's just my opinion. |
# 19:15:29 |
y_morin |
mingwandroid: However, overwriting the .config is probably something we do *not* want. .config is provided by the user, we should not touch that; in any case. |
# 19:15:31 |
mingwandroid |
y_morin: well your opinion matters anyway. |
# 19:15:38 |
y_morin |
mingwandroid: Thank you! :-) |
# 19:16:18 |
y_morin |
mingwandroid: It's 20:15 here, dinner is almost ready, I have to go. Cheers, TTYL... |
# 19:16:29 |
mingwandroid |
ok thanks. i will soon eat too. |
# 19:41:32 |
laeckerv |
quits : Ping timeout: 244 seconds |
# 19:55:47 |
laeckerv |
joins #crosstool-ng |
# 20:00:37 |
laeckerv |
quits : Ping timeout: 265 seconds |
# 21:35:34 |
y_morin |
quits : Remote host closed the connection |
# 21:37:12 |
y_morin |
joins #crosstool-ng |
# 21:44:03 |
diorcety |
quits : Read error: Connection reset by peer |
# 21:50:27 |
y_morin |
mingwandroid: Hello again. I was looking at the bash manual (sonce ct-ng is mostly bash scripts), and there is a better way to pass back values to a caller (not that I suggest we do it, mond you) |
# 21:51:20 |
y_morin |
mingwandroid: In bash, it is posible to declare a variable that is a nameref, that is a reference to abnother variable (like a pointer in C, if you want, or better yet, like a reference in C++) |
# 21:52:50 |
y_morin |
mingwandroid: For example, in a function: local -n ref="${1}"; ref=foo would set the variable passed as first argument to the value 'foo' |
# 21:53:31 |
y_morin |
mingwandroid: That is a bit nicer that the eval-dance, but not much more (e.g. changing the variable that is referenced is a bit convoluted) |
# 22:53:14 |
mingwandroid |
y_morin: oh interesting. |
# 22:54:21 |
mingwandroid |
y_morin: what's the compatability of that? dash for example? I will test actually as MSYS2 has dash. |
# 22:55:17 |
mingwandroid |
y_morin: I have coded it already as I discussed earlier, but I am actually somewhat in agreement with you about repeatability. I might run the idea past Bryan, but I don't mind forcing the user to pass the sha1 in .config. |
# 22:56:51 |
mingwandroid |
y_morin: perhaps a compromise of providing a tool to pre-process a .config that has ref names in it could be comtemplated? |
# 22:59:20 |
y_morin |
mingwandroid: We do not care about comatibilty with dash: ct-ng is nade of bash scripts. |
# 23:00:30 |
y_morin |
mingwandroid: As for a pre-processing: it would not solve the issue: the .config is stil the input, has floating referecnes, and we can't tweak the .config. |
# 23:00:50 |
y_morin |
But I'm off to bed now, it's been a pretty hard day here in .fr |
# 23:01:25 |
y_morin |
Walking all the afternoon until pretty late has been a bit tiring! :-) |
# 23:01:55 |
y_morin |
quits : Quit: Je suis Charlie! |