I think the old keyword info gave two useful pieces of information: 1. list of headers 2. version of the headers
Ideally, I'd like to see: 1. list of headers without version info (generated by compiler, put into .o) 2. complete URL pointing at web-based tree of headers including hash code The hash code would be retrieved from the system at compile time. This one URL would be enough to record the version of any system headers that are used. For application specific headers, they can use #ident however they want. For example, by hand-maintaining the version information. --chris Alexander.Gorshenev at Sun.COM wrote: > On Tue, 16 Jun 2009, James Carlson wrote: > >> Alexander.Gorshenev at Sun.COM writes: >>> So is it something to be addressed later or users just have to live >>> with it from now on? >> >> The 'hg keywords' check will tell you to remove that old SCCS-based >> swill when you're updating files. The general rule is that you fix it >> (by removing it) when you come across it. Old files still have it, >> though. > > That is probably an answer a developer of Solaris would give to a > developer of Solaris. Which I am not. I am from Sun Studio team. And > mostly concerned with the fact that a useful functionality has been > taken out without any substitute. > >> The higher level issue (how does versioning work anyway?) has been >> discussed several times. > > I've read the two mail threads mentioned in "Mercurial Workflow"/"No > ID keywords". The thread where it was decided to drop sccs keywords > only talks and cares about versioning of the modules and other sun built > binaries. What I am interested in is to tell which system header files > were used to produce user's binaries. I want to be able to approximate > user's compilation environment given a binary of a questionable origin. > > You see, we put into binaries (to comment sections and debug sections) > lots of info one can use to reproduce the environment: user's ident > values, versions of the compiler components and the linker. And we used > to put useful ident strings of system headers as well. Not anymore. > > So since #pragma-ident-to-comment-section path is now useless to tell the > header files from the binary, I am looking for any reasonable subsitute. > In theory it would be enough to have hg hash. But something more readable > and in a sence monotoneously increasing would be better. > > thanks, > Alexander > > _______________________________________________ > tools-compilers mailing list > tools-compilers at opensolaris.org
