Re: [tor-dev] #if 0 unused functions?
On Sun, Mar 22, 2015 at 12:12 AM, Sebastian Hahn wrote: > Hi there, > > we have some functions which we never call anywhere. In many cases, it > appears we shouldn't delete them from the source because they "belong" > there - the thing I initially stumbled across was > ed25519_seckey_write_to_file(), for example. But I also don't see why > compiling it and potentially including it in our binary makes any sense. > > Thoughts? I think it's a different answer case-by-case. For the function you mention, it's going to get used by my #12498 patch, so it probably shouldn't get disabled. But for other unused functions, we might be better off removing them entirely, rather than wrapping them away for further use. cheers, -- Nick ___ tor-dev mailing list tor-dev@lists.torproject.org https://lists.torproject.org/cgi-bin/mailman/listinfo/tor-dev
Re: [tor-dev] #if 0 unused functions?
Hi teor, > On 22 Mar 2015, at 14:41, teor wrote: >> Date: Sun, 22 Mar 2015 05:12:03 +0100 >> From: Sebastian Hahn >> To: tor-dev@lists.torproject.org >> >> Hi there, >> >> we have some functions which we never call anywhere. In many cases, it >> appears we shouldn't delete them from the source because they "belong" >> there - the thing I initially stumbled across was >> ed25519_seckey_write_to_file(), for example. But I also don't see why >> compiling it and potentially including it in our binary makes any sense. >> >> Thoughts? >> >> Cheers >> Sebastian > > I'm think that removing / disabling unused code is a great idea, but I wonder > what happens over the long term: > > [How] do we prevent bit-rot on the disabled code? > > Does the disabled code have unit tests? I meant code which doesn't have unit tests currently. > Should we add unit tests or keep them active to prevent bit-rot? > Or should any unit tests be disabled too? > > I wonder if we could unit test unused code, but not link it into the tor > binary. > This would make sure it works if we ever want to use it in future, but > without the risks and performance impacts of including unused code in the > binary. > > There could be issues with this strategy that I haven't thought of. Like the > added maintenance cost involved in keeping unused code and unit tests > up-to-date. > > Perhaps we could: > * remove entirely if we're unlikely to ever use the code, > * disable if we want to keep it around for future reference, and > * unit test it, but not link it in to the tor binary, if we think we'll use > it soon. > > Which category do most of the unused functions fall into? I don't know how much of this code there actually is, I would just do it as I spot such code, not as a big project to identify it all (tho looking at a call graph including unit tests should give you some idea). Cheers Sebastian signature.asc Description: Message signed with OpenPGP using GPGMail ___ tor-dev mailing list tor-dev@lists.torproject.org https://lists.torproject.org/cgi-bin/mailman/listinfo/tor-dev
Re: [tor-dev] #if 0 unused functions?
> Date: Sun, 22 Mar 2015 05:12:03 +0100 > From: Sebastian Hahn > To: tor-dev@lists.torproject.org > > Hi there, > > we have some functions which we never call anywhere. In many cases, it > appears we shouldn't delete them from the source because they "belong" > there - the thing I initially stumbled across was > ed25519_seckey_write_to_file(), for example. But I also don't see why > compiling it and potentially including it in our binary makes any sense. > > Thoughts? > > Cheers > Sebastian I'm think that removing / disabling unused code is a great idea, but I wonder what happens over the long term: [How] do we prevent bit-rot on the disabled code? Does the disabled code have unit tests? Should we add unit tests or keep them active to prevent bit-rot? Or should any unit tests be disabled too? I wonder if we could unit test unused code, but not link it into the tor binary. This would make sure it works if we ever want to use it in future, but without the risks and performance impacts of including unused code in the binary. There could be issues with this strategy that I haven't thought of. Like the added maintenance cost involved in keeping unused code and unit tests up-to-date. Perhaps we could: * remove entirely if we're unlikely to ever use the code, * disable if we want to keep it around for future reference, and * unit test it, but not link it in to the tor binary, if we think we'll use it soon. Which category do most of the unused functions fall into? teor teor2345 at gmail dot com pgp 0xABFED1AC https://gist.github.com/teor2345/d033b8ce0a99adbc89c5 teor at blah dot im OTR C3C57B23 349825DE 929A1DEF C3531C25 A32287ED signature.asc Description: Message signed with OpenPGP using GPGMail ___ tor-dev mailing list tor-dev@lists.torproject.org https://lists.torproject.org/cgi-bin/mailman/listinfo/tor-dev
[tor-dev] #if 0 unused functions?
Hi there, we have some functions which we never call anywhere. In many cases, it appears we shouldn't delete them from the source because they "belong" there - the thing I initially stumbled across was ed25519_seckey_write_to_file(), for example. But I also don't see why compiling it and potentially including it in our binary makes any sense. Thoughts? Cheers Sebastian ___ tor-dev mailing list tor-dev@lists.torproject.org https://lists.torproject.org/cgi-bin/mailman/listinfo/tor-dev
