66367 - Renaming exported internal functions

Status: RESOLVED FIXED

(NSS :: Libraries, defect, P1)

Description

[Wan-Teh Chang]

We have to export some internal functions from
the nss3 shared library for use by the ssl3 or
smime3 shared library.  We do not want our users
to use these internal functions that may change
or even be removed in a future NSS release.

One way to prevent users from using these exported
internal functions is to rename them.  My proposal
is to rename function Foo to __Foo, adding double
underscore in front of the function name.

The current list of functions that we want to rename
are:
CERT_NewTempCertificate;
nss_InitLock;
PK11_CreateContextByRawKey;
PK11_GetKeyData;

CERT_TraversePermCertsForNickname;
CERT_TraversePermCertsForSubject;
CERT_ClosePermCertDB;
CERT_DecodeDERCertificate;
HMAC_Begin;
HMAC_Create;
HMAC_Destroy;
HMAC_Finish;
HMAC_Update;
PBE_CreateContext;
PBE_DestroyContext;
PBE_GenerateBits;

Bob and Nelson think we should modify the pkcs12 code
to not use the HMAC functions.  Then we won't need to
export those HMAC symbols at all.

Comment 1

[Wan-Teh Chang]

Attachment: Proposed patch.

Comment 2

[Wan-Teh Chang]

Attachment: New private header file mozilla/security/nss/lib/nss/nssrename.h.

Comment 3

[Wan-Teh Chang]

Please review my patch and the new nssrename.h header.

Comment 4

[Wan-Teh Chang]

Attachment: Revised patch. (HMAC_ symbols are no longer exported.)

Comment 5

[Wan-Teh Chang]

Attachment: Revised nssrename.h. (HMAC_ symbols are no longer exported.)

Comment 6

[Robert Relyea]

OK, these look good as well.

bob

Comment 7

[Wan-Teh Chang]

I decided to rename "nssrename.h" to "nssrenam.h" to follow
the 8.3 naming convention.

I checked in my latest patch on the tip, with the above change.

Comment 8

[Wan-Teh Chang]

Attachment: A shell script to search for "evil functions" in a source tree.

Comment 9

[Wan-Teh Chang]

I found that some of the "evil functions" are being used
by our clients.  This work is not done.

Comment 10

[Wan-Teh Chang]

All known uses of the "evil" functions in our clients' code
have been removed.  Marked the bug as fixed.