On Sat, Jul 18, 2020 at 09:44:08PM +0200, Klemens Nanni wrote: > On Sat, Jul 18, 2020 at 09:28:50PM +0200, Marc Espie wrote: > > Definitely not a fan... > > I hate squishing pipes in there.... > Why? Due to addmittedly ugly file descriptor handling? > > In that regard, I was also wondering whether our new ksh(1) pipefail > addition could be of use here to simplify things. > > > TEST is okay because dpb doesn't handle it so far. > Would BUILD_LOGFILE interfere with dpb(1)? I have never used it so far. > > > I would hate hate hate to see delayed outputs from other parts > What do you mean with delayed? tee(1)'s output is unbuffered. > > > If you want logs while building manually, well, I did superscript and > > portslogger a while back > Never heard of superscript. > > portslogger(1) is useful but always requires manual usage - it gets in > my way and logs are potentially lost if I forget to start the build with > it in the first place. > > Ideally, I'd like to have logging as built-in part of the framework just > like TEST_LOGFILE is, and not be provided through external tools which > lay outside of the default PATH - working with logs would then only be a > matter of copying/inspecting files *afterwards* if need be, without > having to construct `make | some | pipeline' beforehand.
Definitely seconded. Having that feature by default without having knobs to push (or remembering another external obscure command) would be helpful... the only concern being objdir getting bigger :) Landry
