lld bug: https://bugs.llvm.org/show_bug.cgi?id=44138
--
You are receiving this because you commented.
Reply to this email directly or view it on GitHub:
https://github.com/rpm-software-management/rpm/issues/950#issuecomment-558802296___
Rpm-maint mailin
Fair enough. Fell free to close this. I will forward your analysis to the lld
developers.
--
You are receiving this because you commented.
Reply to this email directly or view it on GitHub:
https://github.com/rpm-software-management/rpm/issues/950#issuecomment-558800517_
Yes, I used ac37755c60ba19103f08f04d07ca8f1d640153d6.
It is impressive how fast sha1 is now, but it is still much slower than xxhash.
Linking a release version of scylla I got
none 2.022252056 seconds
fast 2.061659922 seconds
md5 2.740843349 seconds
sha1 2.318114449 seconds
--
You are receivi
This is the birthday problem, so a good 64 bit hash needs about 32 bit worth of
hashes for a collision. Note that there is precedent for using only 64 bits.
The dwarf standard (http://www.dwarfstd.org/doc/DWARF5.pdf section 7.32 "Type
Signature Computation") uses the last 8 bytes of md5 for typ
Currently rpm requires build ids to be between 16 and 64 bytes in size:
https://github.com/rpm-software-management/rpm/blob/b7d427728b8ba8734ba47d51849a5736bdd727cd/build/files.c#L1883
Since the main intention of build-id is to avoid accidentally using the wrong
binary, lld defaults to xxhash wh
