On Wed, 28 Oct 2026 18:29:13 +0100, Antoine Jacoutot <[email protected]> wrote: > > On Wed, Oct 28, 2026 at 12:20:51PM +0000, Stuart Henderson wrote: > > On 2026/10/28 13:03, Kirill A. Korinsky wrote: > > > On Wed, 28 Oct 2026 12:38:46 +0100, > > > Stuart Henderson <[email protected]> wrote: > > > > > > > > On 2026/10/28 12:30, Theo Buehler wrote: > > > > > On Wed, Oct 28, 2026 at 11:20:49AM +0000, Stuart Henderson wrote: > > > > > > On 2026/10/28 11:48, Kirill A. Korinsky wrote: > > > > > > > > > > > > > > > > Not a fan of this. > > > > > > > > kmos@ added the COMPILER line for sparc64 build at the time. I > > > > > > > > don't know what > > > > > > > > the error was and if it was fixable. > > > > > > > > > > > > > > > > > > > > > > I had talked with kmos@, and he, as expected, hasn't recall nor > > > > > > > have logs of > > > > > > > that error. > > > > > > > > > > > > > > I also checked http://build-failures.rhaalovely.net/sparc64/ and > > > > > > > here to > > > > > > > trace of this. > > > > > > > > > > > > I don't see any error, I checked build logs. > > > > > > > > > > > > libxml-2.10.2 built ok with base compiler on sparc64 on 2022-08-31 > > > > > > > > > > > > The subsequent build on 2022-09-10 was with ports-gcc > > > > > > > > > > > > > > > > kmos often reduced the fallout in package snapshots by fixing things > > > > > up > > > > > before the bulks finished. This is great for users. The downside is > > > > > that > > > > > the logs of the failures were lost because (as I understand it) the > > > > > script only pushes out failures showing up in the summary.log. > > > > > > > > > > > > > aha, there are two builds in that log. errors are in libxml2-py.c; > > > > the python bindings are no longer built. > > > > > > [...] > > > > > > > libxml2-py.c:196: error: expected '=', ',', ';', 'asm' or > > > > '__attribute__' before 'PyObject' > > > > libxml2-py.c:324: error: expected '=', ',', ';', 'asm' or > > > > '__attribute__' before 'PyObject' > > > > libxml2-py.c:1557: error: expected '=', ',', ';', 'asm' or > > > > '__attribute__' before 'PyObject' > > > > libxml2-py.c:1578: error: expected '=', ',', ';', 'asm' or > > > > '__attribute__' before 'PyObject' > > > > libxml2-py.c:2331: error: expected '=', ',', ';', 'asm' or > > > > '__attribute__' before 'PyObject' > > > > libxml2-py.c:2358: error: expected '=', ',', ';', 'asm' or > > > > '__attribute__' before 'PyObject' > > > > libxml2-py.c:3270: error: expected '=', ',', ';', 'asm' or > > > > '__attribute__' before 'PyObject' > > > > libxml2-py.c:3283: error: expected '=', ',', ';', 'asm' or > > > > '__attribute__' before 'PyObject' > > > > libxml2-py.c:3330: error: expected '=', ',', ';', 'asm' or > > > > '__attribute__' before 'PyObject' > > > > libxml2-py.c:3541: error: expected '=', ',', ';', 'asm' or > > > > '__attribute__' before 'PyObject' > > > > libxml2-py.c:6377: error: expected '=', ',', ';', 'asm' or > > > > '__attribute__' before 'PyObject' > > > > libxml2-py.c:6464: error: expected '=', ',', ';', 'asm' or > > > > '__attribute__' before 'PyObject' > > > > libxml2-py.c:6734: error: expected '=', ',', ';', 'asm' or > > > > '__attribute__' before 'PyObject' > > > > libxml2-py.c:6937: error: expected '=', ',', ';', 'asm' or > > > > '__attribute__' before 'PyObject' > > > > libxml2-py.c:7020: error: expected '=', ',', ';', 'asm' or > > > > '__attribute__' before 'PyObject' > > > > libxml2-py.c:13533: error: expected '=', ',', ';', 'asm' or > > > > '__attribute__' before 'PyObject' > > > > Error while executing cc -DHAVE_CONFIG_H -I. -I.. -I../include > > > > -I../include -I/usr/local/include/python3.9 -I/usr/local/include > > > > -I/usr/local/include -O2 -pipe -MT libxml2-py.lo -MD -MP -MF > > > > .deps/libxml2-py.Tpo -c libxml2-py.c -fPIC -DPIC -o .libs/libxml2-py.o > > > > > > Thanks, what perfectly explains why: python needs ports-gcc and expect > > > that > > > python's bindings won't need it is naive, indeed. > > > > > > But libxml port has that comment: > > > > > > # Python bindings are planned to be removed in the 2.16 release; > > > # circular dependency: libxml -> doxygen -> git -> xmlto -> libxml > > > > > > so, it reads like we good with move back to autotools and drop ports-gcc > > > requirements, right? > > > > > > Special that upstream declared C89 as target and it really builds with > > > -std=c89 at CI: > > > https://github.com/GNOME/libxml2/blob/master/.gitlab-ci.yml#L39-L43 > > > > > > without python :) > > > > > > -- > > > wbr, Kirill > > > > > > > It's the general "adding more things on the path to building gcc" that > > I'd prefer to avoid - whichever way it's done it increases the number > > of ports which need extra care when dealing with updates. And since a > > build without ports-gcc is pretty much useless, it's extra work for bulk > > builders (and increases the chance of a bad build getting pushed out to > > ftp if the people signing them don't notice) whenever something breaks > > gcc (the more frontends enabled, the more likely) or one of the ports on > > the path to it. > > > > I am lost :-) > Do whatever you think is best. >
I continue investigating how cobol is made, and I had discovered at least one more missed hard dependency: libiconv; Here missed means that they haven't stated that it is required here: https://gcc.gnu.org/install/prerequisites.html GCC can pick libiconv during build and link libstdc++ agains libiconv, so, something is wrong here. -- wbr, Kirill
