[v2] build: fix soname info for 19.11 compatiblity
diff mbox series

Message ID 20191211151628.1512802-1-bruce.richardson@intel.com
State Superseded
Headers show
Series
  • [v2] build: fix soname info for 19.11 compatiblity
Related show

Checks

Context Check Description
ci/Intel-compilation fail apply issues
ci/travis-robot warning Travis build: failed
ci/iol-mellanox-Performance success Performance Testing PASS
ci/iol-testing success Testing PASS
ci/iol-intel-Performance success Performance Testing PASS
ci/checkpatch warning coding style issues

Commit Message

Bruce Richardson Dec. 11, 2019, 3:16 p.m. UTC
The soname for each stable ABI version should be just the ABI version major
number without the minor number. Unfortunately both major and minor were
used causing version 20.1 to be incompatible with 20.0.

This patch fixes the issue by switching from 2-part to 3-part ABI version
numbers so that we can keep 20.0 as soname and using the final digits to
identify the 20.x releases which are ABI compatible. This requires changes
to both make and meson builds to handle the three-digit version and shrink
it to 2-digit for soname.

Fixes: cba806e07d6f ("build: change ABI versioning to global")

Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
Signed-off-by: Thomas Monjalon <thomas@monjalon.net>
---

V2: 
* adjusted the meson version to work correctly with both 2-part and
  3-part ABI versions, so it will work correctly even with no changes
  for a 21.x ABI version number
* adjusted the versions of the experimental libs so that they are
  consistent between meson and make, and also consistent with v19.11

---
 ABI_VERSION         |  2 +-
 config/meson.build  | 16 +++++++++++++---
 drivers/meson.build |  4 ++--
 lib/meson.build     |  4 ++--
 mk/rte.lib.mk       | 11 +++++++----
 5 files changed, 25 insertions(+), 12 deletions(-)

Comments

David Marchand Dec. 12, 2019, 8:27 a.m. UTC | #1
Hello Bruce,

On Wed, Dec 11, 2019 at 4:16 PM Bruce Richardson
<bruce.richardson@intel.com> wrote:
>
> The soname for each stable ABI version should be just the ABI version major
> number without the minor number. Unfortunately both major and minor were
> used causing version 20.1 to be incompatible with 20.0.
>
> This patch fixes the issue by switching from 2-part to 3-part ABI version
> numbers so that we can keep 20.0 as soname and using the final digits to
> identify the 20.x releases which are ABI compatible. This requires changes
> to both make and meson builds to handle the three-digit version and shrink
> it to 2-digit for soname.
>
> Fixes: cba806e07d6f ("build: change ABI versioning to global")
>
> Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
> Signed-off-by: Thomas Monjalon <thomas@monjalon.net>

There is an issue with the ethtool example.

  INSTALL-APP server
  INSTALL-MAP server.map
cat: /home/dmarchan/dpdk/examples/ethtool/lib/ABI_VERSION: No such
file or directory
  CC rte_ethtool.o
  LD librte_ethtool.so.0.
  INSTALL-LIB librte_ethtool.so.0.
gmake[3]: stat:
/home/dmarchan/builds/i686-native-linux-gcc+shared+debug+default/examples/ethtool/lib/i686-native-linux-gcc/lib/librte_ethtool.so.0.:
Too many levels of symbolic links
== ethtool-app
Ferruh Yigit Dec. 12, 2019, 8:57 a.m. UTC | #2
On 12/12/2019 8:27 AM, David Marchand wrote:
> Hello Bruce,
> 
> On Wed, Dec 11, 2019 at 4:16 PM Bruce Richardson
> <bruce.richardson@intel.com> wrote:
>>
>> The soname for each stable ABI version should be just the ABI version major
>> number without the minor number. Unfortunately both major and minor were
>> used causing version 20.1 to be incompatible with 20.0.
>>
>> This patch fixes the issue by switching from 2-part to 3-part ABI version
>> numbers so that we can keep 20.0 as soname and using the final digits to
>> identify the 20.x releases which are ABI compatible. This requires changes
>> to both make and meson builds to handle the three-digit version and shrink
>> it to 2-digit for soname.
>>
>> Fixes: cba806e07d6f ("build: change ABI versioning to global")
>>
>> Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
>> Signed-off-by: Thomas Monjalon <thomas@monjalon.net>
> 
> There is an issue with the ethtool example.
> 
>   INSTALL-APP server
>   INSTALL-MAP server.map
> cat: /home/dmarchan/dpdk/examples/ethtool/lib/ABI_VERSION: No such
> file or directory
>   CC rte_ethtool.o
>   LD librte_ethtool.so.0.
>   INSTALL-LIB librte_ethtool.so.0.
> gmake[3]: stat:
> /home/dmarchan/builds/i686-native-linux-gcc+shared+debug+default/examples/ethtool/lib/i686-native-linux-gcc/lib/librte_ethtool.so.0.:
> Too many levels of symbolic links
> == ethtool-app
> 
> 

It is linking against itself, in 'examples/ethtool/lib/build/lib':
librte_ethtool.so -> librte_ethtool.so.0.
librte_ethtool.so.0. -> librte_ethtool.so.0.
Bruce Richardson Dec. 12, 2019, 11:44 a.m. UTC | #3
On Thu, Dec 12, 2019 at 08:57:50AM +0000, Ferruh Yigit wrote:
> On 12/12/2019 8:27 AM, David Marchand wrote:
> > Hello Bruce,
> > 
> > On Wed, Dec 11, 2019 at 4:16 PM Bruce Richardson
> > <bruce.richardson@intel.com> wrote:
> >>
> >> The soname for each stable ABI version should be just the ABI version major
> >> number without the minor number. Unfortunately both major and minor were
> >> used causing version 20.1 to be incompatible with 20.0.
> >>
> >> This patch fixes the issue by switching from 2-part to 3-part ABI version
> >> numbers so that we can keep 20.0 as soname and using the final digits to
> >> identify the 20.x releases which are ABI compatible. This requires changes
> >> to both make and meson builds to handle the three-digit version and shrink
> >> it to 2-digit for soname.
> >>
> >> Fixes: cba806e07d6f ("build: change ABI versioning to global")
> >>
> >> Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
> >> Signed-off-by: Thomas Monjalon <thomas@monjalon.net>
> > 
> > There is an issue with the ethtool example.
> > 
> >   INSTALL-APP server
> >   INSTALL-MAP server.map
> > cat: /home/dmarchan/dpdk/examples/ethtool/lib/ABI_VERSION: No such
> > file or directory
> >   CC rte_ethtool.o
> >   LD librte_ethtool.so.0.
> >   INSTALL-LIB librte_ethtool.so.0.
> > gmake[3]: stat:
> > /home/dmarchan/builds/i686-native-linux-gcc+shared+debug+default/examples/ethtool/lib/i686-native-linux-gcc/lib/librte_ethtool.so.0.:
> > Too many levels of symbolic links
> > == ethtool-app
> > 
> > 
> 
> It is linking against itself, in 'examples/ethtool/lib/build/lib':
> librte_ethtool.so -> librte_ethtool.so.0.
> librte_ethtool.so.0. -> librte_ethtool.so.0.

Yes. The issue is that this patch doesn't correct account for external libs
using their own version numbers. The trivial fix for this, which I'll add
in v3 is to make two small changes:

1. Use ?= rather than := when assigning to LIBABIVER in rte.lib.mk
2. Change the LIBABIVER in ethtool/lib to "1.0" rather than "1", since the
code assumes that we have more than a single digit in our version numbers.

Question: Do we need to officially support external libs using our build
system?

* If no (because we assume nobody uses it or otherwise), then we use the two
one-line fixes above and job done.
* If yes, then the makefile logic needs further work to support the case of
having an arbitrary version number. Also, we need to look into the whole
experimental detection logic, as that would probably not apply for external
libs.

/Bruce

PS: No need to ask this question for meson, as there is no assumption that
end user apps use any part of our build systems for their apps - it's
managed through pkg-config only.
Bruce Richardson Dec. 12, 2019, 11:59 a.m. UTC | #4
On Thu, Dec 12, 2019 at 11:44:51AM +0000, Bruce Richardson wrote:
> On Thu, Dec 12, 2019 at 08:57:50AM +0000, Ferruh Yigit wrote:
> > On 12/12/2019 8:27 AM, David Marchand wrote:
> > > Hello Bruce,
> > > 
> > > On Wed, Dec 11, 2019 at 4:16 PM Bruce Richardson
> > > <bruce.richardson@intel.com> wrote:
> > >>
> > >> The soname for each stable ABI version should be just the ABI version major
> > >> number without the minor number. Unfortunately both major and minor were
> > >> used causing version 20.1 to be incompatible with 20.0.
> > >>
> > >> This patch fixes the issue by switching from 2-part to 3-part ABI version
> > >> numbers so that we can keep 20.0 as soname and using the final digits to
> > >> identify the 20.x releases which are ABI compatible. This requires changes
> > >> to both make and meson builds to handle the three-digit version and shrink
> > >> it to 2-digit for soname.
> > >>
> > >> Fixes: cba806e07d6f ("build: change ABI versioning to global")
> > >>
> > >> Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
> > >> Signed-off-by: Thomas Monjalon <thomas@monjalon.net>
> > > 
> > > There is an issue with the ethtool example.
> > > 
> > >   INSTALL-APP server
> > >   INSTALL-MAP server.map
> > > cat: /home/dmarchan/dpdk/examples/ethtool/lib/ABI_VERSION: No such
> > > file or directory
> > >   CC rte_ethtool.o
> > >   LD librte_ethtool.so.0.
> > >   INSTALL-LIB librte_ethtool.so.0.
> > > gmake[3]: stat:
> > > /home/dmarchan/builds/i686-native-linux-gcc+shared+debug+default/examples/ethtool/lib/i686-native-linux-gcc/lib/librte_ethtool.so.0.:
> > > Too many levels of symbolic links
> > > == ethtool-app
> > > 
> > > 
> > 
> > It is linking against itself, in 'examples/ethtool/lib/build/lib':
> > librte_ethtool.so -> librte_ethtool.so.0.
> > librte_ethtool.so.0. -> librte_ethtool.so.0.
> 
> Yes. The issue is that this patch doesn't correct account for external libs
> using their own version numbers. The trivial fix for this, which I'll add
> in v3 is to make two small changes:
> 
> 1. Use ?= rather than := when assigning to LIBABIVER in rte.lib.mk
> 2. Change the LIBABIVER in ethtool/lib to "1.0" rather than "1", since the
> code assumes that we have more than a single digit in our version numbers.
> 
> Question: Do we need to officially support external libs using our build
> system?
> 
> * If no (because we assume nobody uses it or otherwise), then we use the two
> one-line fixes above and job done.
> * If yes, then the makefile logic needs further work to support the case of
> having an arbitrary version number. Also, we need to look into the whole
> experimental detection logic, as that would probably not apply for external
> libs.
> 
> /Bruce
> 
Patch v3 now sent, on the assumption that the answer is "no", or the answer
is "yes, but we can fix that later" :-)
Thomas Monjalon Dec. 12, 2019, 1:41 p.m. UTC | #5
12/12/2019 12:59, Bruce Richardson:
> On Thu, Dec 12, 2019 at 11:44:51AM +0000, Bruce Richardson wrote:
> > On Thu, Dec 12, 2019 at 08:57:50AM +0000, Ferruh Yigit wrote:
> > > On 12/12/2019 8:27 AM, David Marchand wrote:
> > > > Hello Bruce,
> > > > 
> > > > On Wed, Dec 11, 2019 at 4:16 PM Bruce Richardson
> > > > <bruce.richardson@intel.com> wrote:
> > > >>
> > > >> The soname for each stable ABI version should be just the ABI version major
> > > >> number without the minor number. Unfortunately both major and minor were
> > > >> used causing version 20.1 to be incompatible with 20.0.
> > > >>
> > > >> This patch fixes the issue by switching from 2-part to 3-part ABI version
> > > >> numbers so that we can keep 20.0 as soname and using the final digits to
> > > >> identify the 20.x releases which are ABI compatible. This requires changes
> > > >> to both make and meson builds to handle the three-digit version and shrink
> > > >> it to 2-digit for soname.
> > > >>
> > > >> Fixes: cba806e07d6f ("build: change ABI versioning to global")
> > > >>
> > > >> Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
> > > >> Signed-off-by: Thomas Monjalon <thomas@monjalon.net>
> > > > 
> > > > There is an issue with the ethtool example.
> > > > 
> > > >   INSTALL-APP server
> > > >   INSTALL-MAP server.map
> > > > cat: /home/dmarchan/dpdk/examples/ethtool/lib/ABI_VERSION: No such
> > > > file or directory
> > > >   CC rte_ethtool.o
> > > >   LD librte_ethtool.so.0.
> > > >   INSTALL-LIB librte_ethtool.so.0.
> > > > gmake[3]: stat:
> > > > /home/dmarchan/builds/i686-native-linux-gcc+shared+debug+default/examples/ethtool/lib/i686-native-linux-gcc/lib/librte_ethtool.so.0.:
> > > > Too many levels of symbolic links
> > > > == ethtool-app
> > > > 
> > > > 
> > > 
> > > It is linking against itself, in 'examples/ethtool/lib/build/lib':
> > > librte_ethtool.so -> librte_ethtool.so.0.
> > > librte_ethtool.so.0. -> librte_ethtool.so.0.
> > 
> > Yes. The issue is that this patch doesn't correct account for external libs
> > using their own version numbers. The trivial fix for this, which I'll add
> > in v3 is to make two small changes:
> > 
> > 1. Use ?= rather than := when assigning to LIBABIVER in rte.lib.mk
> > 2. Change the LIBABIVER in ethtool/lib to "1.0" rather than "1", since the
> > code assumes that we have more than a single digit in our version numbers.
> > 
> > Question: Do we need to officially support external libs using our build
> > system?
> > 
> > * If no (because we assume nobody uses it or otherwise), then we use the two
> > one-line fixes above and job done.
> > * If yes, then the makefile logic needs further work to support the case of
> > having an arbitrary version number. Also, we need to look into the whole
> > experimental detection logic, as that would probably not apply for external
> > libs.
> > 
> > /Bruce
> > 
> Patch v3 now sent, on the assumption that the answer is "no", or the answer
> is "yes, but we can fix that later" :-)

Yes, that's reasonnable.

Patch
diff mbox series

diff --git a/ABI_VERSION b/ABI_VERSION
index 2e73f8d2a..fcc01369a 100644
--- a/ABI_VERSION
+++ b/ABI_VERSION
@@ -1 +1 @@ 
-20.1
+20.0.1
diff --git a/config/meson.build b/config/meson.build
index 364a8d739..01911ecf9 100644
--- a/config/meson.build
+++ b/config/meson.build
@@ -20,9 +20,19 @@  pver = meson.project_version().split('.')
 major_version = '@0@.@1@'.format(pver.get(0), pver.get(1))
 abi_version = run_command(find_program('cat', 'more'),
 	abi_version_file).stdout().strip()
-# experimental libraries are versioned as 0.majorminor versions, e.g. 0.201
-ever = abi_version.split('.')
-experimental_abi_version = '0.@0@@1@'.format(ever.get(0), ever.get(1))
+
+# Regular libraries have the abi_version as the filename extension
+# and have the soname be all but the final part of the abi_version.
+# Experimental libraries have soname with '0.major'
+# and the filename suffix as 0.majorminor versions,
+# e.g. v20.1 => librte_stable.so.20.1, librte_experimental.so.0.201
+#    sonames => librte_stable.so.20, librte_experimental.so.0.20
+# e.g. v20.0.1 => librte_stable.so.20.0.1, librte_experimental.so.0.2001
+#      sonames => librte_stable.so.20.0, librte_experimental.so.0.200
+abi_va = abi_version.split('.')
+stable_so_version = abi_va.length() == 2 ? abi_va[0] : abi_va[0] + '.' + abi_va[1]
+experimental_abi_version = '0.' + ''.join(abi_va)
+experimental_so_version = '0.' + ''.join(stable_so_version.split('.'))
 
 # extract all version information into the build configuration
 dpdk_conf.set('RTE_VER_YEAR', pver.get(0).to_int())
diff --git a/drivers/meson.build b/drivers/meson.build
index 72eec4608..4b17662b7 100644
--- a/drivers/meson.build
+++ b/drivers/meson.build
@@ -132,10 +132,10 @@  foreach class:dpdk_driver_classes
 
 			if is_experimental != 0
 				lib_version = experimental_abi_version
-				so_version = experimental_abi_version
+				so_version = experimental_so_version
 			else
 				lib_version = abi_version
-				so_version = abi_version
+				so_version = stable_so_version
 			endif
 
 			# now build the static driver
diff --git a/lib/meson.build b/lib/meson.build
index 6ceb5e756..0af3efab2 100644
--- a/lib/meson.build
+++ b/lib/meson.build
@@ -113,10 +113,10 @@  foreach l:libraries
 
 			if is_experimental != 0
 				lib_version = experimental_abi_version
-				so_version = experimental_abi_version
+				so_version = experimental_so_version
 			else
 				lib_version = abi_version
-				so_version = abi_version
+				so_version = stable_so_version
 			endif
 
 			# first build static lib
diff --git a/mk/rte.lib.mk b/mk/rte.lib.mk
index 655a1b143..8ab852a3c 100644
--- a/mk/rte.lib.mk
+++ b/mk/rte.lib.mk
@@ -11,14 +11,16 @@  EXTLIB_BUILD ?= n
 # VPATH contains at least SRCDIR
 VPATH += $(SRCDIR)
 
-ifneq ($(shell grep -s "^DPDK_" $(SRCDIR)/$(EXPORT_MAP)),)
 LIBABIVER := $(shell cat $(RTE_SRCDIR)/ABI_VERSION)
-else ifeq ($(LIBABIVER),)
+SOVER := $(basename $(LIBABIVER))
+ifeq ($(shell grep -s "^DPDK_" $(SRCDIR)/$(EXPORT_MAP)),)
 # EXPERIMENTAL ABI is versioned as 0.major+minor, e.g. 0.201 for 20.1 ABI
-LIBABIVER := 0.$(shell cat $(RTE_SRCDIR)/ABI_VERSION | tr -d '.')
+LIBABIVER := 0.$(shell echo $(LIBABIVER) | tr -d '.')
+SOVER := 0.$(shell echo $(SOVER) | tr -d '.')
 endif
 
 ifeq ($(CONFIG_RTE_BUILD_SHARED_LIB),y)
+SONAME := $(patsubst %.a,%.so.$(SOVER),$(LIB))
 LIB := $(patsubst %.a,%.so.$(LIBABIVER),$(LIB))
 ifeq ($(EXTLIB_BUILD),n)
 CPU_LDFLAGS += --version-script=$(SRCDIR)/$(EXPORT_MAP)
@@ -74,7 +76,7 @@  NO_UNDEFINED := -z defs
 endif
 
 O_TO_S = $(LD) -L$(RTE_SDK_BIN)/lib $(_CPU_LDFLAGS) $(EXTRA_LDFLAGS) \
-	  -shared $(OBJS-y) $(NO_UNDEFINED) $(LDLIBS) -Wl,-soname,$(LIB) -o $(LIB)
+	  -shared $(OBJS-y) $(NO_UNDEFINED) $(LDLIBS) -Wl,-soname,$(SONAME) -o $(LIB)
 O_TO_S_STR = $(subst ','\'',$(O_TO_S)) #'# fix syntax highlight
 O_TO_S_DISP = $(if $(V),"$(O_TO_S_STR)","  LD $(@)")
 O_TO_S_DO = @set -e; \
@@ -133,6 +135,7 @@  $(RTE_OUTPUT)/lib/$(LIB): $(LIB)
 	$(Q)cp -f $(LIB) $(RTE_OUTPUT)/lib
 ifeq ($(CONFIG_RTE_BUILD_SHARED_LIB),y)
 	$(Q)ln -s -f $< $(shell echo $@ | sed 's/\.so.*/.so/')
+	$(Q)ln -s -f $< $(shell echo $@ | sed 's/\.so.*/.so.$(SOVER)/')
 endif
 
 #