IRCloggy #git 2006-09-17

Logs Search ←Prev date Next date→ Channels Documentation

Provider of IRC logs since 2005.
WARNING: As Freenode became unjoinable and lost all warnings in topics, we cannot log channels on Freenode anymore.

2006-09-17

→ anders_ joined00:19
→ GeertB joined00:50
→ anders_ joined00:51
→ ag joined01:29
→ anders_ joined01:29
→ ag joined01:40
→ anders__ joined01:55
→ DrNick joined02:02
→ benlau joined02:19
→ ag joined02:50
→ DrNick joined02:56
→ anders_ joined03:01
→ anders_ joined03:36
→ anders_ joined04:06
→ philv joined05:08
→ devogon joined05:18
→ GeertB joined05:19
→ Gitzilla joined05:20
→ lyakh joined08:27
→ CIA-13 joined08:54
→ robfitz joined09:12
→ chris2 joined09:19
→ _JusSx_ joined09:23
→ somegeek joined09:49
→ ferdy joined10:03
→ GyrosGeier joined10:09
→ coywolf joined10:37
ShadeHawk Gaaah, it seems like vger is filtering out my patches (they didn't appear on git mailing list, yet)11:09
And no reply that "this message is considered spam", or "this host is blacklisted", or something like that11:09
and the email went to Junio, and I got answer... but only my answer to Junio answer to my patch is on mailing list11:10
fonseca You've been very active. It seems every second thread on the mailing list is about gitweb.11:37
.. and my eyes are starting to hurt from looking at all that Perl code. ;)11:39
BTW, have you considered adding diffstat content to gitweb?11:39
ShadeHawk no, for performance reasons. But it would be quite easy to add '--patch-with-stat' to @diff_opts... although we don't have difstat syntax hightlighting as of yet11:46
matled mhhhh11:46
pg=-1...11:46
http://nopaste.php-q.net/24027611:47
> 50 times or so11:47
if ($page =~ m/[^0-9]$/) -> if ($page =~ m/[^0-9]/) probably11:48
ShadeHawk matled: will you send a fix?11:49
matled yes11:49
ShadeHawk TIA11:50
matled: hand editing URL? gitweb is somewhat defensively programmed, but...11:51
matled sure11:52
I read the source and thought 'mh, this check looks weird, why the $ at the end'11:53
looking at all those patches for gitweb perhaps a rewrite would be the best :)11:53
the last PATH_INFO patch is somewhat broken :/11:54
ShadeHawk This ()if ($page =~ m/[^0-9]$/ ) is leftover from original gitweb (as separate project)11:57
it is incrementally being rewritten11:57
matled: what you mean, is broken?11:58
matled well, the simple things are the warnings11:59
how do we handle strings?11:59
is it ok to use them as booleans? like if ($foo) where $foo is a string?11:59
ShadeHawk string is false if it is empty "", or "0", or undef12:00
matled so what about "0"?12:00
ShadeHawk well, if it is undef it is not a string12:00
matled this could be a valid project/filename12:00
ShadeHawk string "0" is equivalent to 0 which is false12:01
matled: corner case12:01
matled well, 0 is a number and "0" is a string which might be a path :)12:01
so we dont care about this?12:01
ShadeHawk and if ($project) is to catch situation when stripping "/" can leave us with empty string12:02
we could use if (defined $foo && length $foo)12:02
or something like that12:02
matled or a custom function for this check (empty($s) is !defined $s || !length $s)12:04
ShadeHawk isempty($s)12:07
→ moh joined12:08
→ philv_ joined12:08
matled is there any difference between $a = undef and undef $a?12:09
ShadeHawk I don't think so. Probably just a question of style.12:10
matled: Fixed warnings in latest patch12:10
(in simplest way possible)12:10
matled i had a patch here too12:11
can you put it somewhere?12:11
ShadeHawk send it too, we could compare them ;-)12:11
"PATCH] gitweb: Fix warnings when PATH_INFO is empty" on git mailing list12:11
matled not yet here12:12
ShadeHawk it looks large bacause of reindent12:14
(and I forgot to include opening [ in the subject... grrr...)12:14
matled reindent?12:14
http://nopaste.php-q.net/24028312:15
ShadeHawk http://nopaste.php-q.net/24028512:16
matled the regexp thing did not complain for me12:16
ShadeHawk it complained for me12:18
Use of uninitialized value in substitution (s///) at gitweb.perl line 277.12:18
PATH_INFO empty (not even /)12:18
matled ah, right12:18
ShadeHawk this path_info work should probably be separated into subroutine, I think12:19
matled I also think this parameter checking should be in a subroutinge12:19
is there any way to use temporary variables there?12:20
ShadeHawk in subroutines?12:20
my $var12:20
matled no, in the parameter checking at the start12:20
ShadeHawk what for?12:20
{ local $var; .... }12:21
matled ah, for example $path_info does not need to be 'our'12:23
no one cares but the short part getting the parameters from it12:23
ShadeHawk in subroutine it can be my $path_info12:26
matled h12:31
mh12:31
you missed the if ($path_info =~ m,^$project/([^/]+)/(.+)$,) part, didn't you?12:31
→ GeertB joined12:31
matled I'll try to make a nice patch with all, I've to patch this part anyway for the export_ok/strict_export thing12:32
→ anders_ joined12:33
→ somegeek joined12:45
matled fatal: ambiguous argument 'git-favicon.png': unknown revision or path not in the working tree.12:48
fun with path_info..12:48
ShadeHawk: more warnings: set p=$project&a=shortlog&h=$branch13:01
Use of uninitialized value in concatenation (.) or string line 287.13:01
ShadeHawk matled: I think not. The whole part with $path_info is in "if ($path_info) { .... }". There are no references to $path_info outside this block13:01
matled and 29913:01
that's what I mean about rewrite, everything I try reveals other problems13:02
mh, that's the old problem13:03
ShadeHawk matled: No cigar. With my path 'p=git.git;a=shortlog;h=next' works O.K.13:04
matled ah, theres the other problem I noted with the path_info patch: validate_input is missing for the branch/file part13:09
_JusSx_ i changed a file but i haven't committed it yet. how can i discard changes? how can i back to version stored to repository?13:13
matled git checkout filename13:13
_JusSx_ matled: thanx13:13
→ anders_ joined13:17
ShadeHawk or, just in case in case, git checkout -- filename13:18
matled ShadeHawk: http://nopaste.php-q.net/240293 any comments? this is on top of master13:18
wget -qO - http://nopaste.php-q.net/down/240293 | git am :)13:19
ShadeHawk Nice.13:21
matled return helps to keep the indention level low :)13:22
ShadeHawk Although the concept itself is flawed, as it assumes that branches are flat. So no "gitweb/xmms2" and "jc/diff" like branches13:22
matled mh, that's still part of the original patch13:23
the whole path_info thing is a bit flawed because it isn't non-ambigous at multiple parts13:23
ShadeHawk well, we could check against the list of branches, but this would seriously affect performance I guess13:24
Extreme case: we could bave branch "ab" and file "sub/zz", and branch "ab/sub" and file "zz" with the new packed refs... ;-)))13:25
with old (current) loose refs we could check if branch with the name being some prefix exists (like we check which part of pathname is repository/project part)13:27
well, with the new, we can use git-show-refs --verify, but it cost some performance13:27
and as of now we can put sha1 in the branch part (as hash), and it would work...13:28
matled we could write the 'backend' (retrieving the data and stuff) in C and only use perl to make the html :)13:29
ShadeHawk matled: But please send your patch if you didn't yet13:29
matled: That is what Git.pm and Git.xs was for... but it appears that git lack suppoer for "repository" abstracting out13:30
matled "mail accepted for delivery"13:30
ShadeHawk Ah.13:30
matled do you send directly from your host to the mailing list? it would be interesting to see what the server says to your mails when they get lost13:31
for example this one got: C="250 2.7.1 Looks like Linux source DIFF email.. BF:<H 0>; S964970AbWIQN3w"13:32
ShadeHawk Your patch is on git mailing list... but you post it as reply to "Re: [PATCH] gitweb: export options" instead of to "[PATCH] gitweb: more support for PATH_INFO based URLs"13:32
No, I send my emails via GMail.13:32
well, patches at least. Articles and replies I usually send via Gmane NNTP (Usenet) interface13:33
I give it a few hours, then send complaint to [email@hidden.address]13:34
matled we could also use the branch:path syntax already supported by git13:45
ShadeHawk Good idea13:46
matled and cat-file -t can tell us what it is13:46
ShadeHawk I'd rather avoid calling git commands (and forking)13:46
matled well, this is once to decide if we do a dir listing or a blob13:47
ShadeHawk Ah, true13:47
by the way, this way we could have "project/filename" syntax, with hash defaulting to "HEAD"13:48
matled mh13:48
and what happens if a branch with name filename comes later?13:49
ShadeHawk git-cat-file -t == git_get_type in gitweb13:49
matled you'd find different things at the same url this way just because of a new branch13:49
which would be a bit strang13:49
ShadeHawk Ooops, there would be problem with directory or filename containing ':'13:50
and it would preclude "project/branch" syntax13:50
matled perhaps project:filename or project/:filename13:50
so if no branch is specified use HEAD13:50
ShadeHawk "project/:filename", i.e. no branch defaults to HEAD, as usual.13:51
matled split path_info at the first :, the first part is repository/branch, the second part is the filename13:51
ShadeHawk No, first find project, then remove project from path_info, then split at first ':' and first part is branch, second is filename13:52
$hash is undef if the branch part is empty13:52
directories leading to project name, and project name itself can contain ':'13:53
only [valid] branches (well, references -- it could be tag name not branch name) couldn't have ':' in them13:54
s/references/refs/13:54
matled yes, but they are quite inconvenient with the git command line tools too :)13:55
http://nopaste.php-q.net/24030413:55
→ ag joined13:56
ShadeHawk I would say "project/branch:/" for directory listing, and check that "project/branch:dir/" works13:57
And rather use only "project/:dir" and "project/:file" as by-the-way shortcuts13:57
$path_info = s/^$project\/?//; my ($branch, $pathname) = split($path_info, ':', 2);14:01
and use "project/: for directory listing of HEAD branch as by-the-way shortcut14:02
Description: matled was too lazy to write a description. ;-))) 14:04
matled http://nopaste.php-q.net/scripts/ makes it usable :)14:08
but most of the time I'm anyway to lazy to find a good description14:10
→ somegeek joined14:11
→ GeertB joined14:17
→ ag joined14:27
→ Gitzilla joined14:28
ShadeHawk matled: Do you subscribe to git mailing list?14:29
matled yes14:29
ShadeHawk No sign of "[PATCH 2/2 (take 3)] gitweb: Make git_get_refs_list do work of git_get_references"?14:32
"take 3" is what is important14:32
matled nope14:32
ShadeHawk almost 5 hours since I've sent it to git mailing list, and no trace...14:32
matled from the message BF gave me for my mail I think mails with patches should get through easily14:33
ShadeHawk http://marc.theaimsgroup.com/?l=git&m=115850281730922&w=2 <-- reply to your http://nopaste.php-q.net/24030414:34
...so I've sent question to postmaster: what happened?...14:34
→ ag joined14:49
pasky So, packing glibc repo from 2.4G took 2.75 hours and the repo size went down to 107M15:03
nice15:03
(--window=50 --depth=25)15:03
chris2 how much with default packing?15:07
GyrosGeier ShadeHawk, somebody set them up the bomb?15:08
pasky how do I know? :)15:08
I will publish the repo15:08
so you can try15:08
chris2 oh no, i got no resources for that ;)15:09
but packs are really nice :)15:09
pasky this packing was done at dualcore P4 on 3GHz with 1G RAM15:12
the cvsimport of the glibc repository though took about two days (over the network)15:12
→ shy joined15:16
← shy left15:17
→ shy joined15:17
pasky [ ] LATEST-IS-2.6.18-rc5 28-Aug-2006 04:10 015:17
[ ] LATEST-IS-2.6.18-rc6 04-Sep-2006 02:25 015:17
[ ] LATEST-IS-2.6.18-rc7 13-Sep-2006 02:15 015:17
wow, multiple choices! ;)15:17
shy hello guys15:17
pasky hi15:17
shy i don't know why gitweb didn't display my project15:17
but projects are in the project_list directory15:18
pasky what does it say?15:20
shy it just say Projects15:21
with Project Description Owner last change15:21
the project_root only contains myproject.git ?15:22
ShadeHawk shy: Up to some patch/commit ago if you wanted hierarchical directory structure, you had to use $projects_list _file_15:23
in the format15:23
path/to/project.git O.+Wner15:24
shy what's O.+Wner ?15:26
philv_ → philv15:29
shy that's don't work :(15:33
→ somegeek1 joined15:34
pasky are you sure you don't have some older output of gitweb cached?15:35
ShadeHawk first, your project has to be under $projectroot (as subdirectory)15:35
shy $projectroot = "/var/cache/git/";15:35
ShadeHawk if it is directly under $projectroot, or if you use latest gitweb, it is enough to have $projects_list unset (make gitweb/gitweb.cgi does that)15:36
or set to $projectroot15:36
shy root@test:/var/cache/git# ls -lh15:36
total 015:36
lrwxrwxrwx 1 root root 31 2006-09-17 15:37 test.git -> /home/shy/git/test/.git15:36
pasky /var/cache is a strange location for $projectroot15:37
shy i know, it's default location of ubuntu package15:37
pasky then ubuntu maintainer is strange ;)15:38
shy: I think you have a permission problem15:38
the webserver can't access /home/shy/git/test/.git15:39
shy permission are right15:40
s15:40
read for everybody15:41
and exe15:41
c15:41
pasky in all the directories all the way down to / ?15:41
shy yes15:43
i use gitweb from ubuntu package15:44
gitweb.cgi is not in http://www.kernel.org/pub/software/scm/gitweb/ ?15:45
pasky nope, it's now part of git15:45
shy i will test with the last package15:47
→ _JusSx_ joined15:52
shy damn dn't work15:54
ShadeHawk erm, you might have the problem I had, namely some directory not having X permission for web server in /home/shy/git/test/.git (i.e. symlink target), and/or .git not having r-x for web server15:54
pasky ShadeHawk: he said he has it all right15:54
so I better hope he does :)15:54
ShadeHawk I jad jnareb in /home/jnareb without x for web server15:54
all right both, or all right in /var/cache/git ?15:55
pasky I've asked about the one in ~15:55
ShadeHawk I had all right in /home/local/scm, (my $projectroot) and it was _not_ enough15:55
pasky shy: try appending ?p=test.git to the url15:55
shy oh great !15:56
that's work15:56
pasky hmm15:57
curious15:57
shy i see the project15:57
but i didn't see commit15:57
only the description, owner etc15:58
pasky what version of git do you have? does git log in the repository run all right?15:58
shy oh16:00
ShadeHawk http://nopaste.php-q.net/24034616:00
shy ?16:01
ShadeHawk: ?16:02
pasky shy: so?16:04
ShadeHawk That is the script I use to debug gitweb (well, after changes, but...)16:04
shy when i go in the commit,i have this message :16:07
403 Forbidden - Unknown commit object.16:07
pasky: the git log is ok16:07
pasky what is your git version?16:08
shy git version 1.1.316:08
pasky ...but gitweb version...?16:09
shy 22016:09
pasky because I'm confused from [17:47] < shy> i will test with the last package16:09
what is "the last package" here?16:10
shy nono16:10
i haven't yet test with the last package :p16:10
pasky well, 1.1.3 is very old16:10
it would be best if you could upgrade both your git and gitweb version16:10
shy i do that16:10
ok so git version 1.4.216:15
i must modify gitweb.cgi directly ?16:17
the new gitweb is not beautiful16:18
pasky what do you mean?16:20
shy maybe the css is not find16:20
have you an exemple of projects_list ?16:21
ShadeHawk git.git Jakub+Narebski16:25
git/git.git Jakub+Narebski16:25
gitweb _might_ need new enough git version. For example newest git requires git-rev-list to have --full-history option, and git to have --git-dir option, which means post 1.4.216:26
old gitweb was self contained.16:27
in new gitweb you must copy gitweb.css, git-logo.png, git-favicon.png and tell gitweb where to find them. Although you can go without16:27
shy uhm i do that16:28
but any change16:28
ShadeHawk Here is the script I use to update gitweb: http://nopaste.php-q.net/24035816:32
with symlinks gitweb in both /var/www/cgi-bin, /var/www/perl, var/www/html to /home/local/gitweb16:33
shy i see in apache log, that it didn't find the css16:35
uhm but he path is right16:35
s/he/the16:36
uhm16:37
gitweb.cgi: Use of uninitialized value in pattern match (m//) at /home/pouik/gitweb/gitweb.cgi line 3416:37
File does not exist: /var/www/home, referer:16:40
...16:40
what's your configuration for apache ?16:41
→ ag joined16:45
shy ok i have correct the error16:46
but i don't know why gitweb got in /var/www/home/16:46
→ pasky joined16:46
→ matled_ joined16:49
shy great !16:53
all works !16:53
thanks guys :p16:53
matled_ → matled16:59
→ anders_ joined17:10
→ boto joined17:17
→ robinr joined17:20
→ DrNick joined17:28
→ devogon joined18:00
→ chexum joined18:01
→ robinr joined18:04
→ devogon joined18:48
→ polyonymous joined19:09
→ DrNick joined19:16
→ xjjk joined19:22
→ GeertB joined19:24
→ lyakh joined19:48
→ dmlb2000 joined20:38
← chexum left20:47
→ robfitz joined22:09
→ segher joined23:38

Logs Search ←Prev date Next date→ Channels Documentation