aboutsummaryrefslogtreecommitdiff
path: root/dpi/cookies.c
AgeCommit message (Collapse)Author
2014-02-02our simplistic TLD list grows more and more worthlesscorvid
2013-09-24our publicsuffix tld list grows even shortercorvid
2012-10-29declarations at start of blockcorvid
2012-10-20cookies dpi tldscorvid
2011-11-11continuedcorvid
2011-11-11fix IPv6 addr recognitioncorvid
I was under the impression that the brackets were part of the host syntax, but a_Url_hostname() shows that they are part of the _authority_ and are stripped out when making the hostname.
2011-11-11locale-independent ASCII character case handlingcorvid
Basically, I and i are different letters in Turkic languages, and this causes problems for str(n)casecmp and toupper/tolower in these locales when dillo is dealing with ASCII.
2011-09-16cookies dpi: MSG only when returning >0 cookiescorvid
2011-09-16update TLDscorvid
What an awful situation it is in general.
2011-05-18don't save huge cookiescorvid
Huge cookies confuse the code that reads them back in. I suppose there's always the possibility of getting a cookies.txt from something else, or a manually-edited one, but... Incidentally, the RFC thinks user agents should allow 4096 bytes for name+value+attributes, which we must be within, say, 50 bytes of.
2011-05-17cookies commentscorvid
I was going to say something in the comments about simplicity and not implementing every little bit of the rfc when we don't know of cookies that make it necessary, but then I suppose that's all implied with dillo.
2011-05-16cookies commentcorvid
2010-08-19turn off some cookie MSGscorvid
2010-05-16disallow nameless cookiescorvid
2010-03-22use the most specific cookiesrc rulecorvid
2010-03-22cookies use host-only as in draft speccorvid
Behaviour should be quite similar to before, but it'll be easier for someone else to understand if it matches the spec's way of doing things more closely.
2010-03-22fix reading maximum expiration date from cookies.txtcorvid
I did check that tm.tm_sec was 0 before adding the max time to it, so that wasn't the problem. (max - 1000) was fine... Surely safer in general to do like this anyway, so I didn't spend time really digging into details this time.
2010-03-22my testcases aren't smart enough for remove_fast on a domain's cookiescorvid
Nothing wrong with it in principle, but...
2010-03-22Cookies_load_cookies()corvid
2010-03-22limit total number of cookiescorvid
2010-03-22cookies some remove_fastcorvid
2010-03-22better handling when too many cookiescorvid
2010-03-21Cookies_delete_nodecorvid
2010-03-21cookies some data structure renamingcorvid
2010-03-10invalid value on invalid monthcorvid
2010-03-10make a point of using -1 for invalid time_tcorvid
2010-03-10if ignoring Expires attr, eat valuecorvid
2010-03-10use tm for max-agecorvid
2010-03-10get rid of the overflowcorvid
2010-03-10separate out server time offsetcorvid
2010-03-10separate out unquotingcorvid
2010-03-10use tm when parsing date stringcorvid
2010-03-10cookies: start using struct tm for time calculationscorvid
Apparently, that's how you're supposed to do things. It's clumsy, but oh well.
2010-03-09another difftimecorvid
2010-03-09cookies: use difftime() for time_t comparisonscorvid
...as Johannes pointed out. (This patch in itself is not meant to fix the time_t overflow problem.)
2010-03-08fix cookie attr leakcorvid
2010-03-04fix dpip cmd leakcorvid
2010-02-02decrease cookies chattinesscorvid
2010-01-24dpi send response when setting cookiesJorge Arellano Cid
2010-01-20cookies improve clock skew handlingcorvid
2010-01-18cookies allow Expires attr's value to be quoted.corvid
sourceforge sends cookies this way. It seems that quotes around values are generally to be ignored, i.e., not stripped or anything, but I asked the draft spec author about this case, and he reports that he has fixed his algorithm to allow leading and trailing delimiters around cookie-date. If I/someone eventually does make the timestamp code follow the algorithm more closely, the stripping would no longer need to be done here. As for whether we'd want to continue to use that code for the stricter server_date that we get from the HTTP Date header, I'm not sure at the moment. I've already added one case to it that wouldn't be legal for the header, though...
2010-01-14Be careful not to delete domain_cookies while still using it.corvid
And don't leak nodes. I never saw any crashing -- and the problem was already in the code before I got to it -- but I was alerted to the matter when puzzling over why there was another round of searching for domain_cookies near the bottom of Cookies_add_cookie.
2010-01-14handle nameless/valueless cookies in cookies.txtcorvid
2010-01-13cookies: be more robust in rejecting IP addr partial matchescorvid
The code was already such that, even if we accepted 123.45 as a domain for host 1.2.123.45, it wouldn't be sent back to anyone. But it would be easy to make some small change later that would break that, so...
2010-01-13cookies fix ipv6 checkcorvid
2010-01-12don't bother to save expired cookiescorvid
2010-01-10cookies update TLDscorvid
2010-01-10cookie handle time overflowcorvid
Jeremy pointed out http://lists.auriga.wearlab.de/pipermail/dillo-dev/2010-January/007144.html that time_t could in principle be a floating type. The cookies dpi assumes that it is an integer type.
2010-01-10make cookies very chatty for nowcorvid
2010-01-09unnecessary testcorvid