Coverity pointed out several of these.

Coverity CID 982331.

Coverify CID 1157776. Removed a superfluous if() that always evaluated
true (and an else clause that never ran), and then re-indented the
function accordingly.

Coverity CID 1215284. The server name is extracted with
Curl_copy_header_value() and passed in to this function, and
copy_header_value can actually can fail and return NULL.

For private keys, use the first match from: user-specified key file
(if provided), ~/.ssh/id_rsa, ~/.ssh/id_dsa, ./id_rsa, ./id_dsa

Note that the previous code only looked for id_dsa files. id_rsa is
now generally preferred, as it supports larger key sizes.

For public keys, use the user-specified key file, if provided.
Otherwise, try to extract the public key from the private key file.
This means that passing --pubkey is typically no longer required,
and makes the key-handling behavior more like OpenSSH.

Coverity CID 1202836. If the proxy environment variable returned an empty
string, it would be leaked. While an empty string is not really a proxy, other
logic in this function already allows a blank string to be returned so allow
that here to avoid the leak.

Coverity CID 1202837. There's a potential risk that 'newurl' gets
overwritten when it was already pointing to allocated memory.

Coverity CID 1215287. There's a potential risk for a memory leak in
here, and moving the free call to be unconditional seems like a cheap
price to remove the risk.

Coverity CID 1215296. There's a potential risk for a memory leak in
here, and moving the free call to be unconditional seems like a cheap
price to remove the risk.

Coverity detected this. CID 1241954. When Curl_poll() returns a negative value
'mcode' was uninitialized. Pretty harmless since this is debug code only and
would at worst cause an error to _not_ be returned...

This patch fixes the "SSL3_WRITE_PENDING: bad write retry" error that
sometimes occurs when sending an email over SMTPS with OpenSSL. OpenSSL
appears to require the same pointer on a write that follows a retry
(CURLE_AGAIN) as discussed here:

http://stackoverflow.com/questions/2997218/why-am-i-getting-error1409f07fssl-routinesssl3-write-pending-bad-write-retr

Mostly because we use C strings and they end at a binary zero so we know
we can't open a file name using an embedded binary zero.

Reported-by:  <research@g0blin.co.uk>

The switch to using Curl_expire_latest() in commit cacdc27f was a
mistake and was against the advice even mentioned in that commit. The
comparison in asyn-thread.c:Curl_resolver_is_resolved() makes
Curl_expire() the suitable function to use.

Bug: http://curl.haxx.se/bug/view.cgi?id=1426
Reported-By: graysky

... to further remove specific TLS backend knowledge from url.c

Previously we did not handle EOF from underlying transport socket and
wrongly just returned error code CURL_AGAIN from http2_recv, which
caused busy loop since socket has been closed.  This patch adds the
code to handle EOF situation and tells the upper layer that we got
EOF.

Bug: http://curl.haxx.se/bug/view.cgi?id=1422
Verified-by: Steve Holme

Removed ISC_REQ_* flags from calls to InitializeSecurityContext to fix
bug in NTLM handshake for HTTP proxy authentication.

NTLM handshake for HTTP proxy authentication failed with error
SEC_E_INVALID_TOKEN from InitializeSecurityContext for certain proxy
servers on generating the NTLM Type-3 message.

The flag ISC_REQ_CONFIDENTIALITY seems to cause the problem according
to the observations and suggestions made in a bug report for the
QT project (https://bugreports.qt-project.org/browse/QTBUG-17322).

Removing all the flags solved the problem.

Bug: http://curl.haxx.se/mail/lib-2014-08/0273.html
Reported-by: Ulrich Telle
Assisted-by: Steve Holme, Daniel Stenberg

Bug: http://curl.haxx.se/mail/lib-2014-09/0064.html

As a sort of step forward, this script will now first try to get the
data from the HTTPS URL using curl, and only if that fails it will
switch back to the HTTP transfer using perl's native LWP functionality.
To reduce the risk of this script being tricked.

Using HTTPS to get a cert bundle introduces a chicken-and-egg problem so
we can't really ever completely disable HTTP, but chances are that most
users already have a ca cert bundle that trusts the mozilla.org site
that this script downloads from.

A future version of this script will probably switch to require a
dedicated "insecure" command line option to allow downloading over HTTP
(or unverified HTTPS).

Test 61 was modified to verify this.

CVE-2014-3620

Reported-by: Tim Ruehsen
URL: http://curl.haxx.se/docs/adv_20140910B.html

By not detecting and rejecting domain names for partial literal IP
addresses properly when parsing received HTTP cookies, libcurl can be
fooled to both send cookies to wrong sites and to allow arbitrary sites
to set cookies for others.

CVE-2014-3613

Bug: http://curl.haxx.se/docs/adv_20140910A.html

Signed-off-by: Catalin Patulea <cat@vv.carleton.ca>

Historically the default "unknown" value for progress.size_dl and
progress.size_ul has been zero, since these values are initialized
implicitly by the calloc that allocates the curl handle that these
variables are a part of.  Users of curl that install progress
callbacks may expect these values to always be >= 0.

Currently it is possible for progress.size_dl and progress.size_ul
to by set to a value of -1, if Curl_pgrsSetDownloadSize() or
Curl_pgrsSetUploadSize() are passed a "size" of -1 (which a few
places currently do, and a following patch will add more).  So
lets update Curl_pgrsSetDownloadSize() and Curl_pgrsSetUploadSize()
so they make sure that these variables always contain a value that
is >= 0.

Updates test579 and test599.

Signed-off-by: Brandon Casey <drafnel@gmail.com>

As the current element in the list is free()d by Curl_llist_remove(),
when the associated connection is pending, reworked the loop to avoid
accessing the next element through e->next afterward.

warning: implicit conversion from enumeration type 'CURLMcode' to
different enumeration type 'CURLcode'

Small follow up to commit 898808fa to use auth constants rather than
hard code value when clearing picked authentication mechanism.

SecCertificateCopyPublicKey() is not available on iPhone. Use
CopyCertSubject() instead to see if the certificate returned by
SecCertificateCreateWithData() is valid.

Reported-by: Toby Peterson

... as the struct is free()d in the end anyway. It was first pointed out
to me that one of the ->msglist assignments were supposed to have been
->pending but was a copy and paste mistake when I realized none of the
clearing of pointers had to be there.

... instead of scanning through all handles, stash only the actual
handles that are in that state in the new ->pending list and scan that
list only. It should be mostly empty or very short. And only used for
pipelining.

This avoids a rather hefty slow-down especially notable if you add many
handles to the same multi handle. Regression introduced in commit
0f147887 (version 7.30.0).

Bug: http://curl.haxx.se/mail/lib-2014-07/0206.html
Reported-by: David Meyer

Forwards the setting as minimum ssl version (if set) to polarssl.  If
the server does not support the requested version the SSL Handshake will
fail.

Bug: http://curl.haxx.se/bug/view.cgi?id=1419