George Rimar via llvm-dev
2016-May-31 13:47 UTC
[llvm-dev] Bug 23231 - support symbol version script / --version-script option
Hi Rui, If you do not have active development on symbol versioning currently, I think I can start working on this, after some more investigations. What do you think ? Best regards, George. -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20160531/cc74537b/attachment.html>
Rafael EspĂndola via llvm-dev
2016-May-31 21:40 UTC
[llvm-dev] Bug 23231 - support symbol version script / --version-script option
On 31 May 2016 at 06:47, George Rimar <grimar at accesssoftek.com> wrote:> Hi Rui, > > > If you do not have active development on symbol versioning currently, > > I think I can start working on this, after some more investigations. > > > What do you think ?Right now I am pretty busy with other bugs, so it would be really nice to have someone working on it. I have upload a reproducible for the freebsd libc to pr23231, where version scripts are the current blocker. The script there is pretty simple. It just has multiple versions and "dependencies", but no globs other than a "local: *". Cheers, Rafael
Rui Ueyama via llvm-dev
2016-Jun-01 22:37 UTC
[llvm-dev] Bug 23231 - support symbol version script / --version-script option
George, If you have bandwidth to work on it, please go ahead. One think I'd note is to start with minimal implementation that is sufficient enough to read and create versioned symbols for minimal version script file. We do not need any wildcard pattern matching (except "local: *", but even this can be implemented later). The initial implementation shouldn't support version hierarchies as well. As to algorithm, implement an algorithm that works O(n) where n is the number of symbols in a version script file. That means you want to iterate over symbols in a version script file instead of interating over symbols in the linker's symbol table. On Tue, May 31, 2016 at 6:47 AM, George Rimar <grimar at accesssoftek.com> wrote:> Hi Rui, > > > If you do not have active development on symbol versioning currently, > > I think I can start working on this, after some more investigations. > > > What do you think ? > > > Best regards, > George. >-------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20160601/cd59c353/attachment.html>
Rafael EspĂndola via llvm-dev
2016-Jun-01 22:46 UTC
[llvm-dev] Bug 23231 - support symbol version script / --version-script option
On 1 June 2016 at 15:37, Rui Ueyama <ruiu at google.com> wrote:> George, > > If you have bandwidth to work on it, please go ahead. One think I'd note is > to start with minimal implementation that is sufficient enough to read and > create versioned symbols for minimal version script file. We do not need any > wildcard pattern matching (except "local: *", but even this can be > implemented later). The initial implementation shouldn't support version > hierarchies as well. > > As to algorithm, implement an algorithm that works O(n) where n is the > number of symbols in a version script file. That means you want to iterate > over symbols in a version script file instead of interating over symbols in > the linker's symbol table.I would also skip matching version in names in the first version (@ver and @@ver). Cheers, Rafael
Reasonably Related Threads
- Should we switch to --hash-style=both by default in LLD ?
- [RFC] - Deduplication of debug information in linkers (LLD).
- [EXTERNAL] Re: preferred way to return expected values
- Bug 26222 - [ELF] wrong functions are called when linking against DSO
- [LLD/ELF] - Should we implement .note.gnu.property and/or Intel CET in LLD ?