Hi James--

Thanks for the prompt followup.

On 03/15/2011 10:50 PM, James Vega wrote:
> This relies on zip properly identifying files as text/binary, which it
> may not do perfectly, as the man page calls out.

Note that zip is not the only tool that upstreams use to build a pkzip
archive.

My experience is that pkzip archivers tend to err on the side of storing
text files as binary (in which case this fix wouldn't make a
difference), not identifying binary files as text (in which case this
fix would actively hurt, as you say).

(and of course, when the archiver doesn't err at all, passing -a is
clearly the right thing to do)

> I'd wager that *nix
> tools are pretty good at handling non-native line endings (at least
> better than my experience with Windows tools), so is it necessary to do
> the conversion?  Have you run into issues with CR/LF line ending files
> that prompted this patch?

sure.  as i mentioned, i'm looking at peter gutmann's cryptlib.  His
manual [0] says:

        To unzip the code under Unix use the -a option to ensure that
        the text files are converted to the Unix format.

If i decompress the archive as upstream expects, and create a diff
against it, that diff will not apply properly against a tarball repacked
by the current incarnation of uscan.

So i'm faced with the choice of:

 0) not using uscan's repacked tarballs, or

 1) distributing patches that don't apply to properly-unpacked upstream
source.

Does this make sense?

        --dkg

[0]  http://www.cryptlib.com/downloads/manual.pdf

Attachment: signature.asc
Description: OpenPGP digital signature

Reply via email to