Skip to contentSkip to conversation

icu 3.6: do_install failed: Segmentation fault

PATCH In review Robert Yang 3 participants 10 messages Jun 27, 2012 – Jul 3, 2012

lore
Share & tools
lore
#0 lore raw

cover icu 3.6: do_install failed: Segmentation fault

The following changes since commit f4b5afe46ff018570c1949653eb6995428827f6f:

sanity.bbclass: Increase LAYER_CONF_VERSION to match bblayers change (2012-06-25 17:20:06 +0100)

are available in the git repository at:
git://git.pokylinux.org/poky-contrib robert/icu
http://git.pokylinux.org/cgit.cgi/poky-contrib/log/?h=robert/icu

Robert Yang (2):
icu 3.6: dropp apply=yes from path
icu 3.6: do_install failed: Segmentation fault

 .../icu/files/larger-cmd-size.patch                |   27 ++++++++++++++++++++
 meta/recipes-support/icu/icu-3.6.inc               |    9 ++++--
 meta/recipes-support/icu/icu_3.6.bb                |    2 +-
 3 files changed, 34 insertions(+), 4 deletions(-)
 create mode 100644 meta/recipes-support/icu/files/larger-cmd-size.patch
#1 lore raw

PATCH 1/2 icu 3.6: dropp apply=yes from path

The "apply=yes" doesn't appear to be necessary, so drop it.

[YOCTO #2664]

---
 meta/recipes-support/icu/icu-3.6.inc |    8 +++++---
 1 files changed, 5 insertions(+), 3 deletions(-)
meta/recipes-support/icu/icu-3.6.inc +5 -3
index 23447d3..d3391fe 100644
@@ -3,11 +3,13 @@ DESCRIPTION = "The International Component for Unicode (ICU) is a mature, portab
 HOMEPAGE = "http://www-01.ibm.com/software/globalization/icu/index.jsp"
 
 BASE_SRC_URI = "ftp://ftp.software.ibm.com/software/globalization/icu/3.6/icu4c-3_6-src.tgz \
-           file://gccfix.patch;apply=yes \
+           file://gccfix.patch \
            file://fix-parallel-build.patch \
-           file://use-g++-for-linking.patch;apply=yes"
+           file://use-g++-for-linking.patch \
+           "
 SRC_URI = "${BASE_SRC_URI} \
-           file://noldlibpath.patch;apply=yes"
+           file://noldlibpath.patch \
+           "
 SRC_URI_virtclass-native = "${BASE_SRC_URI}"
 
 LICENSE = "ICU"
Signature · 1 line
1.7.1
#2 lore raw

PATCH 2/2 icu 3.6: do_install failed: Segmentation fault

In reply to Robert Yang

There is a "Segmentation fault" error when the tmpdir is longer than 470 (or less), this is because it used "char cmd[1024]" which is not enough for the command line. Allocate a larger memory size should fix this problem.

[YOCTO #2664]

---
 .../icu/files/larger-cmd-size.patch                |   27 ++++++++++++++++++++
 meta/recipes-support/icu/icu-3.6.inc               |    1 +
 meta/recipes-support/icu/icu_3.6.bb                |    2 +-
 3 files changed, 29 insertions(+), 1 deletions(-)
 create mode 100644 meta/recipes-support/icu/files/larger-cmd-size.patch
meta/recipes-support/icu/files/larger-cmd-size.patch new file +27 -0
new file mode 100644
index 0000000..4b5f73d
@@ -0,0 +1,27 @@
+Allocate a larger memory size for cmd
+
+The length of the command line can be longer than 1024 sometimes,
+which will cause a "Segmentation fault" error.
+
+Signed-off-by: Robert Yang <[email protected]>
+
+Upstream-Status: Pending
+---
+ tools/pkgdata/pkgdata.c |    2 +-
+ 1 file changed, 1 insertion(+), 1 deletion(-)
+
+diff --git a/tools/pkgdata/pkgdata.c b/tools/pkgdata/pkgdata.c
+--- a/tools/pkgdata/pkgdata.c
++++ b/tools/pkgdata/pkgdata.c
+@@ -439,7 +439,7 @@ main(int argc, char* argv[]) {
+ /* POSIX - execute makefile */
+ static int executeMakefile(const UPKGOptions *o)
+ {
+-    char cmd[1024];
++    char cmd[BUFSIZ];
+     /*char pwd[1024];*/
+     const char *make;
+     int rc;
+-- 
+1.7.10.2
+
meta/recipes-support/icu/icu-3.6.inc +1 -0
index d3391fe..d969caa 100644
@@ -6,6 +6,7 @@ BASE_SRC_URI = "ftp://ftp.software.ibm.com/software/globalization/icu/3.6/icu4c-
            file://gccfix.patch \
            file://fix-parallel-build.patch \
            file://use-g++-for-linking.patch \
+           file://larger-cmd-size.patch \
            "
 SRC_URI = "${BASE_SRC_URI} \
            file://noldlibpath.patch \
meta/recipes-support/icu/icu_3.6.bb +1 -1
index d7be522..f941acf 100644
@@ -1,6 +1,6 @@
 require icu-3.6.inc
 
-PR = "r8"
+PR = "r9"
 
 SRC_URI[md5sum] = "6243f7a19e03e05403ce84e597510d4c"
 SRC_URI[sha256sum] = "5135e8d69d6206d320515df7aeee7027711ab1aef9d8dbf29571a97a9746b041"
Signature · 1 line
1.7.1
#3 lore raw
Quoted 31 lines · Robert Yang

There is a "Segmentation fault" error when the tmpdir is longer than 470 (or less), this is because it used "char cmd[1024]" which is not enough for the command line. Allocate a larger memory size should fix this problem.

[YOCTO #2664]

---
  .../icu/files/larger-cmd-size.patch                |   27 ++++++++++++++++++++
  meta/recipes-support/icu/icu-3.6.inc               |    1 +
  meta/recipes-support/icu/icu_3.6.bb                |    2 +-
  3 files changed, 29 insertions(+), 1 deletions(-)

create mode 100644 meta/recipes-support/icu/files/larger-cmd-size.patch

meta/recipes-support/icu/files/larger-cmd-size.patch new file +11 -0
new file mode 100644
index 0000000..4b5f73d
@@ -0,0 +1,27 @@
+Allocate a larger memory size for cmd
+
+The length of the command line can be longer than 1024 sometimes,
+which will cause a "Segmentation fault" error.
+
+Signed-off-by: Robert Yang<[email protected]>
+
+Upstream-Status: Pending
+---
+ tools/pkgdata/pkgdata.c |    2 +-
+ 1 file changed, 1 insertion(+), 1 deletion(-)

meta/recipes-support/icu/files/larger-cmd-size.patch line 21

Quoted 10 lines
meta/recipes-support/icu/files/larger-cmd-size.patch new file +10 -0
+
+diff --git a/tools/pkgdata/pkgdata.c b/tools/pkgdata/pkgdata.c
+--- a/tools/pkgdata/pkgdata.c
++++ b/tools/pkgdata/pkgdata.c
+@@ -439,7 +439,7 @@ main(int argc, char* argv[]) {
+ /* POSIX - execute makefile */
+ static int executeMakefile(const UPKGOptions *o)
+ {
+-    char cmd[1024];
++    char cmd[BUFSIZ];

I've changed the cmd[BUFSIZ] to cmd[2048], which should be enough for us, and pushed it to the git repo again:

git://git.pokylinux.org/poky-contrib robert/icu
http://git.pokylinux.org/cgit.cgi/poky-contrib/log/?h=robert/icu

// Robert

Quoted 30 lines
meta/recipes-support/icu/files/larger-cmd-size.patch new file +6 -0
+     /*char pwd[1024];*/
+     const char *make;
+     int rc;
+--
+1.7.10.2
+
meta/recipes-support/icu/icu-3.6.inc +1 -0
index d3391fe..d969caa 100644
@@ -6,6 +6,7 @@ BASE_SRC_URI = "ftp://ftp.software.ibm.com/software/globalization/icu/3.6/icu4c-
             file://gccfix.patch \
             file://fix-parallel-build.patch \
             file://use-g++-for-linking.patch \
+           file://larger-cmd-size.patch \
             "
  SRC_URI = "${BASE_SRC_URI} \
             file://noldlibpath.patch \
meta/recipes-support/icu/icu_3.6.bb +1 -1
index d7be522..f941acf 100644
@@ -1,6 +1,6 @@
  require icu-3.6.inc
 
-PR = "r8"
+PR = "r9"
 
  SRC_URI[md5sum] = "6243f7a19e03e05403ce84e597510d4c"
  SRC_URI[sha256sum] = "5135e8d69d6206d320515df7aeee7027711ab1aef9d8dbf29571a97a9746b041"
#4 lore raw
Quoted 3 lines · Robert Yang

I've changed the cmd[BUFSIZ] to cmd[2048], which should be enough for us, and
pushed it to the git repo again:

ideally allocating a string on heap and then freeing it after use would fix this issue once for all.

#5 lore raw
Quoted 8 lines · Khem Raj
Quoted 3 lines · Robert Yang

I've changed the cmd[BUFSIZ] to cmd[2048], which should be enough for us, and
pushed it to the git repo again:

ideally allocating a string on heap and then freeing it after use would fix this issue once for all.

Thanks Khem, but the problem is that even we use:

void *malloc(size_t size)

We should still use a constant size since the length of the command is unknown (or not obviously known):

#else
    sprintf(cmd, "%s %s%s -f %s %s %s %s %s",
        make,
        o->install ? "INSTALLTO=" : "",
        o->install ? o->install    : "",
        o->makeFile,
        o->clean   ? "clean"      : "",
        o->rebuild ? "rebuild"    : "",
        o->install ? "install"    : "",
        o->makeArgs);

So I simply increase the size of cmd[] which is a simple fix.

// Robert

Quoted 6 lines

_______________________________________________
Openembedded-core mailing list
[email protected]
http://lists.linuxtogo.org/cgi-bin/mailman/listinfo/openembedded-core

#6 lore raw
Quoted 2 lines · Robert Yang

We should still use a constant size since the length of the command is unknown (or not obviously known):

thats the point. Cant you use srtlen(cmd) ?

#7 lore raw
Quoted 6 lines · Khem Raj
Quoted 2 lines · Robert Yang

We should still use a constant size since the length of the command is unknown (or not obviously known):

thats the point. Cant you use srtlen(cmd) ?

I'm afraid not, as the code shows below, we should strlen() each component of the cmd:

    sprintf(cmd, "%s %s%s -f %s %s %s %s %s",
        make,
        o->install ? "INSTALLTO=" : "",
        o->install ? o->install    : "",
        o->makeFile,
        o->clean   ? "clean"      : "",
        o->rebuild ? "rebuild"    : "",
        o->install ? "install"    : "",
        o->makeArgs);

and this is only the part for linux os, it still has similar parts for U_WINDOWS and OS400 (the component are different), if we strlen() each one of them, that would make it complicated. the icu pkg does use many malloc() in the other part of code, but it uses the char array here, I think that maybe this is the reason.

// Robert

Quoted 1 line
#8 lore raw
Quoted 28 lines · Robert Yang
Quoted 10 lines · Khem Raj
Quoted 3 lines · Robert Yang

We should still use a constant size since the length of the command is unknown (or not obviously known):

thats the point. Cant you use srtlen(cmd) ?

I'm afraid not, as the code shows below, we should strlen() each component of the cmd:

   sprintf(cmd, "%s %s%s -f %s %s %s %s %s",
       make,
       o->install ? "INSTALLTO=" : "",
       o->install ? o->install    : "",
       o->makeFile,
       o->clean   ? "clean"      : "",
       o->rebuild ? "rebuild"    : "",
       o->install ? "install"    : "",
       o->makeArgs);

Quoted 6 lines

and this is only the part for linux os, it still has similar parts for U_WINDOWS and OS400 (the component are different), if we strlen() each one of them, that would make it complicated. the icu pkg does use many malloc() in the other part of code, but it uses the char array here, I think that maybe this is the reason.

OK its fine what you have

Quoted 4 lines

// Robert

Quoted 1 line
#9 lore raw
In reply to Robert Yang
Quoted 16 lines · Robert Yang

The following changes since commit f4b5afe46ff018570c1949653eb6995428827f6f:

   sanity.bbclass: Increase LAYER_CONF_VERSION to match bblayers change (2012-06-25 17:20:06 +0100)

are available in the git repository at:
git://git.pokylinux.org/poky-contrib robert/icu
http://git.pokylinux.org/cgit.cgi/poky-contrib/log/?h=robert/icu

Robert Yang (2):
icu 3.6: dropp apply=yes from path
icu 3.6: do_install failed: Segmentation fault

  .../icu/files/larger-cmd-size.patch                |   27 ++++++++++++++++++++
  meta/recipes-support/icu/icu-3.6.inc               |    9 ++++--
  meta/recipes-support/icu/icu_3.6.bb                |    2 +-
  3 files changed, 34 insertions(+), 4 deletions(-)
Quoted 9 lines

create mode 100644 meta/recipes-support/icu/files/larger-cmd-size.patch

_______________________________________________
Openembedded-core mailing list
[email protected]
http://lists.linuxtogo.org/cgi-bin/mailman/listinfo/openembedded-core

Merged into OE-Core the latest version with 2048

Thanks
Sau!