198622 - redundant calls to memset in freebl

Status: RESOLVED FIXED

(NSS :: Libraries, enhancement, P3)

Description

[Ian McGreer]

The function alg_fips186_1_x3_1 zeroes a SHA1Context, then calls SHA1_Begin,
which does the same thing.  The first call should be removed.

Comment 1

[Ian McGreer]

Attachment: remove redundant call to memset

Comment 2

[Wan-Teh Chang]

Comment on attachment 118039 [details] [diff] [review]
remove redundant call to memset

r=wtc.

Comment 3

[Ian McGreer]

Morphing this bug.  There are apparently multiple locations where redundant
memset calls are made in freebl.  Might as well handle them all here.

The next two are in MD5_NewContext and SHA1_NewContext.  Both use PORT_ZAlloc. 
However, before the contexts can be used, MD5_Begin and SHA1_Begin must be
called.  Both of those functions use memset/memcpy/assignment to initialize all
of the context's fields.  So the ZAlloc in NewContext is redundant.

Comment 4

[Ian McGreer]

Attachment: remove more redundant memsets

Note that in sha512.c, the NewContext functions do not use ZNew, so only
sha_fast.c needed the change.

Comment 5

[Wan-Teh Chang]

Comment on attachment 118067 [details] [diff] [review]
remove more redundant memsets

r=wtc. Please change _Begin to SHA1_Begin and MD5_Begin when
you check it in.

Comment 6

[Nelson Bolyard (seldom reads bugmail)]

Comment on attachment 118067 [details] [diff] [review]
remove more redundant memsets

I've reviewed these changes.  r=nbb.  Ian, maybe you should ask Wan-Teh to
superreview them.

Comment 7

[Nelson Bolyard (seldom reads bugmail)]

Comment on attachment 118067 [details] [diff] [review]
remove more redundant memsets

I think wtc and I both marked this bug r= about the same time. sr=nelsonb

Comment 8

[Ian McGreer]

Thanks for the reviews.  Checked in; marking fixed.