compilation bug fix for bsds
Yarin
yarin at warpmail.net
Sat Jan 15 13:52:45 EST 2011
Looks like the attachment got scrubbed again.
Here's the patch again, hopefully what ever it is that's causing your lines to get wrapped ~70 chars long won't hurt it.
--- ./parse/Csv2_rr_soa.c 2010-08-28 17:13:20.000000000 -0500
+++ ./parse/Csv2_rr_soa.c 2011-01-15 12:17:46.293201632 -0600
@@ -34,17 +34,18 @@
*/
int csv2_b4_at(int32 in) {
- /* [0-9a-zA-Z\-\_\+\%\!\^\=] */
+ /* [0-9a-zA-Z\-\_\+\%\!\^\=\.] */
return (csv2_is_alphanum(in) || in == '+' || in == '%' ||
- in == '!' || in == '^' || in == '=');
+ in == '!' || in == '^' || in == '=' || in == '.');
}
-/* Process an address in the form 'a at foo.bar.baz.' or 'a.foo.bar.baz.' */
+/* Process an address in the form 'a at foo.bar.baz.', 'a.foo.bar.baz.', or 'a\.b\.c at foo.bar.baz.' */
js_string *process_mbox(csv2_read *stream) {
js_string *o;
int32 look;
int x;
+ unsigned int prescaped = 0;
o = process_1stchar(stream,csv2_is_alphanum_ordot,"Z");
if(o == 0) {
@@ -69,13 +70,18 @@
js_destroy(o);
return 0;
}
- if(look == '@' || look == '.') {
- if(csv2_append_utf8(o, look) < 0) {
+ if(look == '@' || (look == '.' && !prescaped)) {
+ if(prescaped) {
+ csv2_error(stream,"You can't escape an @ or use it in the local part in mbox");
+ js_destroy(o);
+ return 0;
+ }
+ if(csv2_append_utf8(o, '@') < 0) { // use an '@' regardless, to support local part '.'s
csv2_error(stream,"Error appending character");
js_destroy(o);
return 0;
}
- if(look == '.') {
+ if(look == '.') { // note: this block is asking for trouble if js_append_dname() is well behaved
look = csv2_read_unicode(stream);
if(csv2_is_text(look)) {
if(csv2_append_utf8(o, look) < 0) {
@@ -93,12 +99,21 @@
}
break;
}
- if(csv2_b4_at(look)) {
+ if(look == '\\') {
+ if(prescaped) {
+ csv2_error(stream,"Unexpected character before @ in mbox, backslashes are illegal");
+ js_destroy(o);
+ return 0;
+ }
+ prescaped = 1; // the parser will forgive escaping of regular characters without complaining
+ }
+ else if(csv2_b4_at(look)) {
if(csv2_append_utf8(o, look) < 0) {
csv2_error(stream,"Error appending character");
js_destroy(o);
return 0;
}
+ prescaped = 0;
}
else {
csv2_error(stream,"Unexpected character before @"
----- Original message -----
From: "Yarin" <yarin at warpmail.net>
To: list at maradns.org
Date: Sat, 15 Jan 2011 12:47:02 -0600
Subject: Re: compilation bug fix for bsds
> Then again, there is a reason I prefer "-u" diffs.
Yes, and so I see why.
> Yes, '@' should do the right thing with a SOA email address
Oh, '@' does do the right thing, it's just that the parser doesn't accept '.'s in the local part of the email address, because when it sees one, it thinks your trying to use it in place of the '@'.
I've attached a proposed patch (released under the two-clause BSD license) that will fix this, by treating escaped dots differently. So, for example, the address "a.b.c at foo.bar.baz" can be expressed with "a\.b\.c.foo.bar.baz." This is the way the dig utility does it.
Yarin
----- Original message -----
From: "Sam Trenholme" <strenholme.usenet at gmail.com>
To: list at maradns.org
Date: Sat, 15 Jan 2011 01:44:59 -0700
Subject: Re: compilation bug fix for bsds
> Haha, it looks like you've updated duende since your latest official
> release, so when you added the patch, one of the patch's lines got
> stuffed in a new comment. I've confirmed that everything else is
> where it's suppose to be though. The below diff, applied on top of it,
> should put it where it goes.
Thanks for catching that. I had a cold this last weekend which forced
me to rest a lot and was spending too much time on Usenet [1] when I
should have been looking closer to your patch. Then again, there is a
reason I prefer "-u" diffs.
I have fixed it:
http://samiam.org/blog/20110115.html
> Just recently, MaraDNS freaked when I tried to use a dot in the local
> part of an SOA record. I then discovered that MaraDNS holds to the
> classic, using the first dot in the address instead of an at sign. Would
> you accept a patch that works with this to support dots in the local
> part? The address would be parsed the same way dig prints it, that
> is, with the local dots escaped, when lacking an at sign.
You know, I have learned a lot with implementing that code. Like the
wisdom for using an interpreter (or a meta-compiler like yacc/bison)
for parsing text. Especially a fairly complex parser like the one
used for csv2 zone file parsing. And, while I'm at it, the wisdom of
making substantial changes to the parser (the optional use of '~' to
separate DNS records in MaraDNS 1.3 so it is easier to convert BIND
zone files in to CSV2 zone files) once it is written.
Yes, '@' should do the right thing with a SOA email address in
MaraDNS. If it doesn't, it's a bug.
- Sam
[1] Usenet was the protocol defining the internet until the web took
over in the mid-1990s. Usenet was the place to share useful and cool
things until the people sharing useful information moved on. I
checked it out again this last week to relive a bit of nostalgia;
Usenet, alas, is dead.
More information about the list
mailing list