Tue Mar 7 23:13:19 ART 2006 Alberto Bertogli tagged 0.15 Tue Mar 7 23:13:01 ART 2006 Alberto Bertogli * Version 0.15. Thu Feb 23 22:43:41 ART 2006 Alberto Bertogli tagged 0.15-rc1 Thu Feb 23 22:37:00 ART 2006 Alberto Bertogli * Fix repoencoding, this time for real (I hope). Thu Feb 23 21:54:19 ART 2006 Alberto Bertogli * Fix to allow having a single repoencoding. Thu Feb 23 21:31:28 ART 2006 Alberto Bertogli * Rewrite encoding handling. This patch rewrites fixu8() and affects a couple of places where decode() was used directly. It has the benefit of removing the ugly previous function, replacing it with... well, another ugly function, but with a different kind of uglyness. It also supports multiple encodings, so if the first one fails, a second one is tried (and so on). There still are some corner cases with file named in alternative encodings, but should work much better than the older code. Thu Feb 23 16:47:31 ART 2006 Alberto Bertogli * Include the repository in log_times() output. Thu Feb 23 15:51:21 ART 2006 Alberto Bertogli * Quote filenames in several places. There were several places using filenames in URLs without quoting them properly, this fixes them. Thanks to VMiklos for the report. Thu Feb 23 15:34:51 ART 2006 Alberto Bertogli * Filter '"' in filenames. For security, don't allow '"' in filenames. If there is high demand, some alternative workaround could be implemented. Thu Feb 23 14:21:30 ART 2006 Alberto Bertogli * Add /etc/darcsweb to the module lookup path. This allows darcsweb's configuration to be in /etc/darcsweb instead of the current directory. It's mostly wanted by distributions which want to isolate the config file to keep things clean. A similar patch was sent by Gaetan Lehmann (Mandriva), VMiklos (Frugalware) and Fabian Linzberger (Debian). Wed Feb 22 01:47:15 ART 2006 Alberto Bertogli * Move some one-time used imports to the user location. As a minor but noticeable optimization to darcsweb's load time, move the importing of sha and email.Utils modules to the place where they're used (only one in both cases). This takes the time spent on imports from 0.027s to 0.021s, which will probably be slightly noticeable on slower machines, reducing the time taken by a cache hit from 0.030s to 0.024s (on misses it doesn't make a difference). Tue Feb 21 18:24:30 ART 2006 Alberto Bertogli * Begin with alt = True in listings. I think it looks better if we start a listing with alt = True (the dark one). Mon Feb 20 00:23:19 ART 2006 Alberto Bertogli * Implement time logging. This record adds an option to log the time it took darcsweb to present a page. Unfortunately, because of the way it's implemented, it's impossible to isolate the time it takes to call darcs; however, every darcs invocation is logged too, to allow independant measures. Sun Feb 19 14:42:23 ART 2006 Alberto Bertogli * Change cache file permissions 600. Change cache file permissions to 600. This should work on Windows too because the bits are ignored, according to python's documentation. Thu Jan 12 20:52:34 ART 2006 Alberto Bertogli * Fix typos in the BOLA license. Mon Jan 9 19:25:04 ART 2006 Alberto Bertogli tagged 0.14 diff -rN -u old-darcsweb/config.py.sample new-darcsweb/config.py.sample --- old-darcsweb/config.py.sample 2006-03-07 23:24:59.000000000 -0300 +++ new-darcsweb/config.py.sample 2006-03-07 23:24:59.000000000 -0300 @@ -46,6 +46,11 @@ # set the limit to a very high number, they will take time. #searchlimit = 100 + # If you want to log the times it took darcsweb to present a page, + # uncomment this option. The value should be a file writeable by + # darcsweb. + #logtimes = "/tmp/darcsweb_times" + # # From now on, every class is a repo configuration, with the same format @@ -75,6 +80,9 @@ # like 'utf-8' or 'UTF8') if you expect darcsweb to work properly. # This is because to workaround a bug in darcs we need to do some # codec mangling and it needs special cases for UTF8. + # You can, optionally, specify multiple encodings; they're tried in + # order, and if one fails to decode a string, the next one is tried. + # Example: repoencoding = "utf8", "latin1" repoencoding = "latin1" # as with the base configuration, the footer is also optional, and it diff -rN -u old-darcsweb/darcsweb.cgi new-darcsweb/darcsweb.cgi --- old-darcsweb/darcsweb.cgi 2006-03-07 23:24:59.000000000 -0300 +++ new-darcsweb/darcsweb.cgi 2006-03-07 23:24:59.000000000 -0300 @@ -8,25 +8,33 @@ and Christian Gierke """ +import time +time_begin = time.time() import sys import os import string -import time import stat -import sha import cgi import cgitb; cgitb.enable() import urllib import xml.sax from xml.sax.saxutils import escape as xml_escape -import email.Utils +time_imports = time.time() - time_begin iso_datetime = '%Y-%m-%dT%H:%M:%SZ' +# In order to be able to store the config file in /etc/darcsweb, it has to be +# added to sys.path. It's mainly used by distributions, which place the +# default configuration there. +sys.path.append('/etc/darcsweb') + # empty configuration class, we will fill it in later depending on the repo class config: pass +# list of run_darcs() invocations, for performance measures +darcs_runs = [] + # exception handling def exc_handle(t, v, tb): try: @@ -58,7 +66,7 @@ def filter_file(s): - if '..' in s: + if '..' in s or '"' in s: raise 'FilterFile FAILED' if s == '/': return s @@ -80,42 +88,24 @@ # I _hate_ this. def fixu8(s): - openpos = s.find('[_') - if openpos < 0: - # small optimization to avoid the conversion to utf8 and - # entering the loop - if type(s) == unicode: - # workaround for python < 2.4 - return s.encode('utf8') - else: - return s.decode(config.repoencoding).encode('utf8') + """Calls _fixu8(), which does the real work, line by line. Otherwise + we choose the wrong encoding for big buffers and end up messing + output.""" + n = [] + for i in s.split('\n'): + n.append(_fixu8(i)) + return string.join(n, '\n') + +def _fixu8(s): + if type(s) == unicode: + return s.encode('utf8', 'replace') + for e in config.repoencoding: + try: + return s.decode(e).encode('utf8', 'replace') + except UnicodeDecodeError: + pass + raise 'DecodingError', config.repoencoding - s = s.encode(config.repoencoding).decode('raw_unicode_escape') - while openpos >= 0: - closepos = s.find('_]', openpos) - if closepos < 0: - # not closed, probably just luck - break - - # middle should be something like 'c3', so we get it by - # removing the first three characters ("[_\") - middle = s[openpos + 3:closepos] - if len(middle) == 2: - # now we turn middle into the character "\xc3" - char = chr(int(middle, 16)) - - # finally, replace s with our new improved string, and - # repeat the ugly procedure - char = char.decode(config.repoencoding) - mn = '[_\\' + middle + '_]' - s = s.replace(mn, char, 1) - openpos = s.find('[_', openpos + 1) - - if config.repoencoding != 'utf8': - s = s.encode('utf8') - else: - s = s.encode('raw_unicode_escape', 'replace') - return s def escape(s): s = xml_escape(s) @@ -233,6 +223,36 @@ realf = filter_file(config.repodir + '/' + fname) return realf +def log_times(cache_hit, repo = None, event = None): + if not config.logtimes: + return + + time_total = time.time() - time_begin + processing = time_total - time_imports + if not event: + event = action + if cache_hit: + event = event + " (hit)" + s = '%s\n' % event + + if repo: + s += '\trepo: %s\n' % repo + + s += """\ + total: %.3f + processing: %.3f + imports: %.3f\n""" % (time_total, processing, time_imports) + + if darcs_runs: + s += "\truns:\n" + for params in darcs_runs: + s += '\t\t%s\n' % params + s += '\n' + + lf = open(config.logtimes, 'a') + lf.write(s) + lf.close() + # # generic html functions @@ -245,7 +265,7 @@ -