Created attachment 345102 [details] Example input The attached example comes from the configure script of dev-lang/swi-prolog-6.3.13. If decoded correctly, it will result in an executable Java class file of 311 bytes. app-arch/sharutils-4.13.4 will instead create a file of 318 bytes, with extra '\n' introduced in various places, thus breaking the file format. As the issue appears to be fixed in current git HEAD, I assume that the removal of the TRY_PUTCHAR line in http://git.savannah.gnu.org/cgit/sharutils.git/diff/src/uudecode.c?id=262472c10eef9d12660934be8fe509040aac7965 is the fix. In that case the issue was introduced by http://git.savannah.gnu.org/cgit/sharutils.git/diff/src/uudecode.c?id=3e286f0822bce050a22ed3d8f9d9478600575479 which apparently is a commit AFTER the 4.13.3 release, so this is a regression and 4.13.3 is fine.
Created attachment 345118 [details, diff] Proposed fix I can confirm that removing the TRY_PUTCHAR line in question fixes this issue. I therefore suggest revbumping the ebuild and including the attached patch. I also contacted upstream via the bug-gnu-utils mailing list, suggesting a release including this fix, but so far my message is apparently still awaiting moderation.
Interestingly, db 4.8.30 runs uudecode in its ./configure, which never used to produce correct output. uudecode from sharutils 4.13.4 is bad for this, as usual. uudecode from sharutils 4.13.3 produces *correct* output. So, I kinda confirm ;)
I bit off too much and I'm going to be away for two weeks, so a released fix won't happen until May. Until then, please either drop back or pull from the repo. Thank you.
Alternatively: http://autogen.sourceforge.net/data/sharutils-4.13.5pre4.tar.xz
(In reply to comment #4) i think we can wait for 4.13.5. i doubt this impacts many people ;).
4.13.5 added to CVS.
