autotools: Attempt to clean build directory when rerunning configure
PATCH 1/1 autotools: Attempt to clean build directory when rerunning configure
Existing logic in autotools_preconfigure() will clean the build directory before rerunning configure in the case where ${S} != ${B}. It is, however, desirable to make at least some attempt to clean the build artifacts for in-tree builds since many packages lack the appropriate dependency rules to force a recompile if (for example) CFLAGS has changed. Making distclean would be better but is not practical because many packages don't support it properly. Even "make clean" is not universally available, so we ignore any failure from that command and just carry on regardless.
--- meta/classes/autotools.bbclass | 13 +++++++++---- 1 file changed, 9 insertions(+), 4 deletions(-)
meta/classes/autotools.bbclass +9 -4
@@ -93,10 +93,15 @@ CONFIGURESTAMPFILE = "${WORKDIR}/configure.sstate" autotools_preconfigure() { if [ -n "${CONFIGURESTAMPFILE}" -a -e "${CONFIGURESTAMPFILE}" ]; then - if [ "`cat ${CONFIGURESTAMPFILE}`" != "${BB_TASKHASH}" -a "${S}" != "${B}" ]; then - echo "Previously configured separate build directory detected, cleaning ${B}" - rm -rf ${B} - mkdir ${B} + if [ "`cat ${CONFIGURESTAMPFILE}`" != "${BB_TASKHASH}" ]; then + if [ "${S}" != "${B}" ]; then + echo "Previously configured separate build directory detected, cleaning ${B}" + rm -rf ${B} + mkdir ${B} + else + echo "Previously configured source directory detected, cleaning ${S}" + ( oe_runmake clean ) || true + fi fi fi }
Signature · 1 line
1.7.10.4
Quoted 9 lines · Phil Blundell
Existing logic in
autotools_preconfigure()will clean the build directory before rerunning configure in the case where ${S} != ${B}. It is, however, desirable to make at least some attempt to clean the build artifacts for in-tree builds since many packages lack the appropriate dependency rules to force a recompile if (for example) CFLAGS has changed. Making distclean would be better but is not practical because many packages don't support it properly. Even "make clean" is not universally available, so we ignore any failure from that command and just carry on regardless.
Quoted 4 lines
Signed-off-by Phil Blundell <[email protected]>--- meta/classes/autotools.bbclass | 13 +++++++++---- 1 file changed, 9 insertions(+), 4 deletions(-)
How extensively has this been tested? I have a suspicion that we might run into a number of issues doing this throughout the tree :(
I'm wondering how many recipes will suffer from issues with "make clean" verses how many recipes would struggle if we defaulted ${B} != ${S}?
Cheers,
Richard
Quoted 24 lines
meta/classes/autotools.bbclass +9
-4@@ -93,10 +93,15 @@ CONFIGURESTAMPFILE = "${WORKDIR}/configure.sstate" autotools_preconfigure() { if [ -n "${CONFIGURESTAMPFILE}" -a -e "${CONFIGURESTAMPFILE}" ]; then - if [ "`cat ${CONFIGURESTAMPFILE}`" != "${BB_TASKHASH}" -a "${S}" != "${B}" ]; then - echo "Previously configured separate build directory detected, cleaning ${B}" - rm -rf ${B} - mkdir ${B} + if [ "`cat ${CONFIGURESTAMPFILE}`" != "${BB_TASKHASH}" ]; then + if [ "${S}" != "${B}" ]; then + echo "Previously configured separate build directory detected, cleaning ${B}" + rm -rf ${B} + mkdir ${B} + else + echo "Previously configured source directory detected, cleaning ${S}" + ( oe_runmake clean ) || true + fi fi fi }
Quoted 2 lines · Richard Purdie
How extensively has this been tested? I have a suspicion that we might run into a number of issues doing this throughout the tree :(
I've had this in my tree for a few weeks and the only package I've found so far that went wrong was polkit, see separate patch for that. (There a few other packages for which "make clean" fails in a harmless manner.) But it's true that I haven't built the whole of world this way yet.
Quoted 3 lines
I'm wondering how many recipes will suffer from issues with "make clean" verses how many recipes would struggle if we defaulted ${B} != ${S}?
My guess is that "make clean" is slightly more likely to work (or, at least, not break anything) on average. But defaulting ${B} != ${S} would definitely be a desirable thing in the longer term and I would certainly be in favour of moving in that direction.
p.