The reference on SHA(-0) is FIPS 180 (archived scan) of 1993 May 11, referenced on the NIST website (the link recently added there is to another scan lacking page 1 and the one before, thus foreword, abstract, explanation..). There's also this text draft of 1992 January 22 (also here with less indication on origin).
In Section 7 of  (page 9), the line which reads
(b) For t=16 to 79 let Wt = Wt-3 XOR Wt-8 XOR Wt-14 XOR Wt-16.
is to be replaced by
(b) For t=16 to 79 let Wt = S1(Wt-3 XOR Wt-8 XOR Wt-14 XOR Wt-16).
where S1 is a left circular shift by one bit as defined in Section 3 of  (page 6), namely S1(X) = (X << 1) v (X >> 31).
Note: What's kept of the original PfC has messed-up sub/superscripts (a modern web artifact), and misses the closing parenthesis on the modified expression for Wt (possibly an omission in the original).
This is was a technical change in the design, critical for improved security, rather than a correction of a typographical error as could be understood by casual reading of the PfC. This change was introduced without rationale, and initially without ensuing changes in the test vectors. It is thus reasonable to suspect that it was intentionally sneaked quietly, with knowledge that the original was not as secure as intended against educated collision attacks.
Changing source code for SHA-1 into one for SHA(-0) is easy: just remove the one-bit rotations of a 32-bit word of the 512-bit message block. For example, in the reference implementation of RFC 6234, change
W[t] = SHA1_ROTL(1, W[t-3] ^ W[t-8] ^ W[t-14] ^ W[t-16]);
W[t] = ( W[t-3] ^ W[t-8] ^ W[t-14] ^ W[t-16]);
Here is a copy directly from NIST at the DOI site.
And yes, I simply asked and got multiple, very polite responses. Thanks NIST!
You can try this - https://github.com/tyage/sha0/blob/master/lib/sha0.rb I don't know how good or accurate it is.