Age | Commit message (Collapse) | Author |
|
|
|
|
|
Note: use strncasecmp including the \0 ending to exact compare the strings
|
|
Last-Update: 2016-06-18
During builds against glibc version 2.23, an error report mentioned
that implicit conversion of nil to an empty string is not automatic.
|
|
Bug-Debian: http://bugs.debian.org/818852
http://bugs.debian.org/831270
Last-Update: 2016-09-02
When built against glibc version 2.23, the provided tests
fail completely. The cause turns out to be a segmentation
fault in the rule interpreter. The upstream code implicitly
depends on a counter field being zeroed at return from malloc.
Instead use calloc() for the large structures.
|
|
|
|
|
|
|
|
Removing UTF8 unbreakable space from ruby file fixed it!
|
|
|
|
Dear Julien Viard de Galbert,
in tests for an updated Debian packaging of netsed
I found myself needing replacement rules that have
an effect in a single direction only, like
s/ae/oe/i s/oe/ae/o
for handling "in coeli" phonetically. Patch is offered.
Regards
Mats Erik Andersson
>From 413cf2d636e16055f732baf359844cf30b18902d Mon Sep 17 00:00:00 2001
From: Mats Erik Andersson <gnu@gisladisker.se>
Date: Sun, 14 Jul 2013 13:24:48 +0200
Subject: [PATCH] Allow rules to depend on direction.
Implement a selector to be appended to rules,
that restrict their applicability to incoming,
or to outgoing packets.
|
|
Dear Julien Viard de Galbert,
running lintian on netsed-1.1 produces two identical
spelling errors. In fact, both are errors in grammar,
since in English "information" is an uncountable noun.
Best regards,
Mats Erik Andersson
>From 64ca85a1f7065c093aee720fd88582e298c2dd6e Mon Sep 17 00:00:00 2001
From: Mats Erik Andersson <debian@gisladisker.se>
Date: Fri, 5 Jul 2013 15:45:02 +0200
Subject: [PATCH] Spelling in binary.
|
|
|
|
|
|
The full help in now printed out on options error (or --help), a brief
note gets out in other error cases suggesting the -h option.
|
|
Without the new test, the listening socket is single stacked since the
computation used "af" instead of "family".
|
|
|
|
Surrounded by macros in case the getopt_long GNU extension is not
available...
|
|
|
|
Using explicit arrays for rules parameters instead on implicit arrays
(build from all remaining parameters).
This will allow to add optional parameters.
|
|
|
|
|
|
When the remote host is targeted using IPv6, and
'net.ipv6.bindv6only=0' holds, then the local
listening socket will accept both address families.
As this intrudes on the available socket name space,
an additional test can be used to reset the IPv6
socket to be single stacked.
|
|
A rule of unlimited validity must be ot the form
`s/this/that', without a final delimiting dash.
This will confuse traditionalists like myself,
expecting 's/this/that/' to deliver the same effect.
.
Thus make a final dash optional, like common sed-syntax.
|
|
Added GPL v2+ in files, added optional Ruby license to ruby files also.
|
|
|
|
The unconditional setting of LINUX_NETFILTER causes FTBFS for
GNU/kfreebsd.
|
|
|
|
|
|
|
|
fakeroot and base directory as suggested by Mats Erik Andersson.
|
|
TTL are counter per connections
|
|
|
|
|
|
|
|
|
|
basically tcp does not need timeout, and for udp the select timeout is adjusted
to the nearest pseudo connection timeout so that we never exits select for no
good reason ;)
|
|
Removed now unused SIGCHLD handler.
Using sigaction() to set SIGINT handler as recommended by signal man page.
|
|
Mostly comment cleanup.
Clean exit on error (closes sockets)
|
|
|
|
structure and function documentation done.
|
|
|
|
both are now using a yaml parsable format, just in case...
|
|
|
|
|
|
udp was slightly brocken: client2server_sed is only for tcp...
refactoring the new connection creation code to share most code
|
|
Do I really deserve it...
|
|
applied the same select mechanism to tcp
this broke the TTL (by connection) feature as fork was silently duplicating
the rule array.
some factoring done.
|
|
|
|
|