RE: gmake-4.4.1 has a performance regression: at least the science/nwchem project now builds much slower
Date: Fri, 29 Mar 2024 16:21:58 UTC
Yuri <yuri_at_FreeBSD.org> wrote on Date: Fri, 29 Mar 2024 06:40:05 UTC : > When the port devel/gmake was updated from 4.3 to 4.4.1 on 2024-03-03, > the science/nwchem port began to build much slower. It used to build in > 1-1.5 hours, and now it builds in 10+ hours and times out on builders. > > Some GNU make changes caused this. > > Did anyone else observe performance problems? I compared a before/after log pairs from 140amd64-default and saw a 3 more port differences. I'll list installing lines for all found: [140amd64-default-job-??] Installing gmake-4.3_3... vs. [140amd64-default-job-??] Installing gmake-4.4.1... [140amd64-default-job-??] | | | `-- Installing libxcb-1.15_2... vs. [140amd64-default-job-??] | | | `-- Installing libxcb-1.16.1... [140amd64-default-job-??] | | | `-- Installing libXdmcp-1.1.3... vs. [140amd64-default-job-??] | | | `-- Installing libXdmcp-1.1.5... [140amd64-default-job-??] | | | | `-- Installing libunistring-1.1... vs. [140amd64-default-job-??] | | | | `-- Installing libunistring-1.2... (The "??"'s are my substitutions to make the lines compared be not automatically different.) I've no specific evidence about which might contribute. For all I know gmake may be all that does. But I figured that noting the other variances might be of use. === Mark Millard marklmi at yahoo.com