On Sat Aug 17, 2024 at 3:52 PM CEST, Max Krummenacher wrote:
> Hi Luca
>
> On Sat, Aug 17, 2024 at 11:48 AM Luca Weiss <luca.weiss@xxxxxxxxxxxxx> wrote:
> >
> > On Thu Jul 25, 2024 at 3:20 PM CEST, max.oss.09 wrote:
> > > From: Max Krummenacher <max.krummenacher@xxxxxxxxxxx>
> > >
> > > conmakehash uses getenv("abs_srctree") from the environment to strip
> > > the absolute path from the generated sources.
> > > However since commit e2bad142bb3d ("kbuild: unexport abs_srctree and
> > > abs_objtree") this environment variable no longer gets set.
> > > Instead use basename() to indicate the used file in a comment of the
> > > generated source file.
> > >
> > > Fixes: 3bd85c6c97b2 ("tty: vt: conmakehash: Don't mention the full path of the input in output")
> > > Signed-off-by: Max Krummenacher <max.krummenacher@xxxxxxxxxxx>
> > >
> > > ---
> > >
> > > drivers/tty/vt/conmakehash.c | 20 +++++++-------------
> > > 1 file changed, 7 insertions(+), 13 deletions(-)
> > >
> > > diff --git a/drivers/tty/vt/conmakehash.c b/drivers/tty/vt/conmakehash.c
> > > index dc2177fec715..82d9db68b2ce 100644
> > > --- a/drivers/tty/vt/conmakehash.c
> > > +++ b/drivers/tty/vt/conmakehash.c
> > > @@ -11,6 +11,8 @@
> > > * Copyright (C) 1995-1997 H. Peter Anvin
> > > */
> > >
> > > +#include <libgen.h>
> > > +#include <linux/limits.h>
> >
> > Hi Max,
> >
> > Not sure this is the best place to ask but this <linux/limits.h> include
> > appears to rely on this file already being installed in /usr/include and
> > is not taken from the Linux source tree that's being built.
> >
> > This mostly manifests in building Linux kernel e.g. in Alpine Linux
> > package build if 'linux-headers' package is not being explicitly
> > installed, failing with
> >
> > drivers/tty/vt/conmakehash.c:15:10: fatal error: linux/limits.h: No such file or directory
> > 15 | #include <linux/limits.h>
> > | ^~~~~~~~~~~~~~~~
> > compilation terminated.
> >
> > Apparently this is (understandably) also a problem when building on
> > macOS:
> > https://lore.kernel.org/all/20240807-macos-build-support-v1-11-4cd1ded85694@xxxxxxxxxxx/
> >
> > I did try that linked patch a bit ago, but unfortunately didn't fix it
> > for the Alpine Linux build environment.
>
> This is a bug I introduced.
>
> Masahiro Yamada already fixed it with [1].
Great, I see it's also in torvalds' tree already, so I assume it'll hit
stable soon as well.
Thanks!
>
> Sorry about that.
>
> Regards
> Max
>
> [1] https://lore.kernel.org/all/20240809160853.1269466-1-masahiroy@xxxxxxxxxx/
>
> >
> > Any ideas?
> >
> > Regards
> > Luca
> >
> >
> > > #include <stdio.h>
> > > #include <stdlib.h>
> > > #include <sysexits.h>
> > > @@ -76,8 +78,8 @@ static void addpair(int fp, int un)
> > > int main(int argc, char *argv[])
> > > {
> > > FILE *ctbl;
> > > - const char *tblname, *rel_tblname;
> > > - const char *abs_srctree;
> > > + const char *tblname;
> > > + char base_tblname[PATH_MAX];
> > > char buffer[65536];
> > > int fontlen;
> > > int i, nuni, nent;
> > > @@ -102,16 +104,6 @@ int main(int argc, char *argv[])
> > > }
> > > }
> > >
> > > - abs_srctree = getenv("abs_srctree");
> > > - if (abs_srctree && !strncmp(abs_srctree, tblname, strlen(abs_srctree)))
> > > - {
> > > - rel_tblname = tblname + strlen(abs_srctree);
> > > - while (*rel_tblname == '/')
> > > - ++rel_tblname;
> > > - }
> > > - else
> > > - rel_tblname = tblname;
> > > -
> > > /* For now we assume the default font is always 256 characters. */
> > > fontlen = 256;
> > >
> > > @@ -253,6 +245,8 @@ int main(int argc, char *argv[])
> > > for ( i = 0 ; i < fontlen ; i++ )
> > > nuni += unicount[i];
> > >
> > > + strncpy(base_tblname, tblname, PATH_MAX);
> > > + base_tblname[PATH_MAX - 1] = 0;
> > > printf("\
> > > /*\n\
> > > * Do not edit this file; it was automatically generated by\n\
> > > @@ -264,7 +258,7 @@ int main(int argc, char *argv[])
> > > #include <linux/types.h>\n\
> > > \n\
> > > u8 dfont_unicount[%d] = \n\
> > > -{\n\t", rel_tblname, fontlen);
> > > +{\n\t", basename(base_tblname), fontlen);
> > >
> > > for ( i = 0 ; i < fontlen ; i++ )
> > > {
> >
