Commit c40d2bc0c9b60c43b641ac4669c7b8bbcd6134c5 has a problem
at removing the parentheses in parser_parse_responseline.
Difference of pointers was used with intention, don't cast
them separately.
Previous change broke the feature. The error of unique_service_name
in ssdp_request_type should be ignored.
This reverts commit 5944960e172a797a9fcc196291f4046cafa7f6ec.
src/genlib/net/sock.c: In function ‘sock_read_write’:
src/genlib/net/sock.c:172:4: warning: conversion to ‘long int’ from
‘size_t’ may change the sign of the result
There was a problem in HDR_ACCEPT_LANGUAGE case.
It may read from TmpBuf larger amount than allocated,
since condition was always true.
Terminate RespInstr->AcceptLanguageHeader correctly.
Skip allocation if there is already sufficient buffer.
Do not compile most of service_table.c and client_table.c if
--disable-gena is used.
Do not compile urlconfig.c if --disable-webserver is used.
Adding new UPNP_HAVE_xxx variables in upnpconfig.h and upnpconfig.h.in.
Submitted: Yoichi NAKAYAMA ( yoichi ) - 2012-03-08 10:18:39 PST
97a17ff5add73c97844e2fa74456bab4df0800f1 commit breaks build on
windows/msvc since there is no snprintf.
Note:
* Some existing sources use _snprintf when WIN32 is defined, but its
behavior is a bit different from C99 snprintf.
* snprintf does terminate the buffer, so the commit (use buffer size
minus 1 as argument) changes the behavior at the boundary.
* Truncation might be better than crash in some cases. But it may
result in not good.
Submitted: Marcelo Roberto Jimenez ( mroberto ) - 2012-03-08 12:38:57 PST
src/api/upnpapi.c: In function ‘UpnpUnSubscribeAsync’:
src/api/upnpapi.c:2060:6: warning: ‘retVal’ may be used uninitialized in this function
Submitted: Fabrice Fontaine ( ffontaine ) - 2012-03-06 07:36:08 PST
Call to strcpy should be replaced by call to memset and strncpy to
avoid getting buffer overflows.
Submitted: Nick Leverton ( leveret ) - 2010-04-19 07:44:10 PDT
Details: The recent codebase merge has removed a significant API call
which is used by several pupnp devices such as mediatomb and gmediaserver.
UpnpSetVirtualDirCallbacks() has been replaced by individual routines to
set each callback. Essentially this means that 1.6.7 will in fact be a majo
bump and 1.6.6 devices can no longer link against it. Could we have the call
reinstated please, perhaps as a wrapper around the individual calls ? As
it is, all distros will have to patch their 1.6.x apps, rebuild and re-link them.
The other removed API calls and external variables don't seem to be used
by any of the apps I have copies of, but UpnpSetVirtualDirCallbacks is
important for maintaining compatibility within 1.6.x.
Submitted: Sunil ( sunilangadi ) - 2011-10-02 08:28:47 PDT
Details: I observed crash in the below mentioned log statement in
function upnpfinish(file: upnpapi.c).
UpnpPrintf(UPNP_INFO, API, __FILE__, __LINE__, "Exiting UpnpFinish:
UpnpSdkInit is :%d:\n", UpnpSdkInit);
In particular it was crashing in ithread_self in
UpnpDisplayFileAndLine(file upnpdebug.c) on WIN32.
Moving the call ithread_cleanup_library() below the upnp printf call
mentioned above in function upnpfinish fixed the crash but I couldn't get
to the root of the problem.
The problem was observed on WIN32.
Submitted: Fabrice Fontaine ( ffontaine ) - 2012-03-05 06:42:18 PST
gIF_IPV4, gIF_IPV6 and gIF_IPV6_ULA_GUA might be not null terminated.
Moreover, gIF_IPV4 should be 16 characters (INET_ADDRSTRLEN) and not 22
and gIF_IPV6 should be 46 characters (INET6_ADDRSTRLEN) and not 65.
