diff --git a/R-package/DESCRIPTION b/R-package/DESCRIPTION
index fabf5733..becd7bf5 100644
--- a/R-package/DESCRIPTION
+++ b/R-package/DESCRIPTION
@@ -1,8 +1,8 @@
 Package: abess
 Type: Package
 Title: Fast Best Subset Selection
-Version: 0.4.8
-Date: 2023-09-19
+Version: 0.4.9
+Date: 2024-09-09
 Authors@R: c(
   person(given = "Jin", family = "Zhu", email = "zhuj37@mail2.sysu.edu.cn", role = c("aut", "cre"), comment = c(ORCID = "0000-0001-8550-5822")),
   person(given = "Zezhi", family = "Wang", email = "homura@mail.ustc.edu.cn", role = c("aut")),
diff --git a/R-package/NEWS.md b/R-package/NEWS.md
index b7c4d4e4..f3dec772 100644
--- a/R-package/NEWS.md
+++ b/R-package/NEWS.md
@@ -1,3 +1,8 @@
+# abess 0.4.9
+* Fix bug in Cpp level
+* Fix error in: https://www.stats.ox.ac.uk/pub/bdr/clang19/abess.log
+* Fix notes in https://cran.r-project.org/web/checks/check_results_abess.html
+
 # abess 0.4.8
 * Support no-intercept GLM model by param 'fit.intercept'.
 * Allow to restrict the range of estimation for beta by param 'beta.high' and 'beta.low'.
diff --git a/R-package/R/abess.R b/R-package/R/abess.R
index b6aa0513..5a5d8052 100644
--- a/R-package/R/abess.R
+++ b/R-package/R/abess.R
@@ -174,10 +174,10 @@ abess <- function(x, ...) UseMethod("abess")
 #' 
 #' \code{abess} support some well-known advanced statistical methods to analyze data, including 
 #' \itemize{
-#'   \item{sure independent screening: } {helpful for ultra-high dimensional predictors (i.e., \eqn{p \gg n}). Use the parameter \code{screening.num} to retain the marginally most important predictors. See Fan et al (2008) for more details. }
-#'   \item{best subset of group selection: } {helpful when predictors have group structure. Use the parameter \code{group.index} to specify the group structure of predictors. See Zhang et al (2021) for more details. }
-#'   \item{\eqn{l_2} regularization best subset selection: } {helpful when signal-to-ratio is relatively small. Use the parameter \code{lambda} to control the magnitude of the regularization term.}
-#'   \item{nuisance selection: } {helpful when the prior knowledge of important predictors is available. Use the parameter \code{always.include} to retain the important predictors.}
+#'   \item sure independent screening: helpful for ultra-high dimensional predictors (i.e., \eqn{p \gg n}). Use the parameter \code{screening.num} to retain the marginally most important predictors. See Fan et al (2008) for more details. 
+#'   \item best subset of group selection: helpful when predictors have group structure. Use the parameter \code{group.index} to specify the group structure of predictors. See Zhang et al (2021) for more details. 
+#'   \item \eqn{l_2} regularization best subset selection: helpful when signal-to-ratio is relatively small. Use the parameter \code{lambda} to control the magnitude of the regularization term.
+#'   \item nuisance selection: helpful when the prior knowledge of important predictors is available. Use the parameter \code{always.include} to retain the important predictors.
 #' }
 #' The arbitrary combination of the four methods are definitely support. 
 #' Please see [online vignettes](https://abess-team.github.io/abess/articles/v07-advancedFeatures.html) for more details about the advanced features support by \code{abess}. 
diff --git a/R-package/R/coef.abesspca.R b/R-package/R/coef.abesspca.R
index 7c5f066f..34c74954 100644
--- a/R-package/R/coef.abesspca.R
+++ b/R-package/R/coef.abesspca.R
@@ -9,12 +9,12 @@
 #' the coefficient fitted at given \code{support.size}.
 #' If \code{support.size = NULL}, then all coefficients would be returned.
 #' Default: \code{support.size = NULL}.
-#' This parameter is omitted if {sparse.type = "kpc"}.
+#' This parameter is omitted if \code{sparse.type = "kpc"}.
 #' @param kpc An integer vector specifies
 #' the coefficient fitted at given principal component.
 #' If \code{kpc = NULL}, then all coefficients would be returned.
 #' Default: \code{kpc = NULL}.
-#' This parameter is omitted if {sparse.type = "fpc"}.
+#' This parameter is omitted if \code{sparse.type = "fpc"}.
 #'
 #' @param sparse A logical value, specifying whether the coefficients should be
 #' presented as sparse matrix or not. Default: \code{sparse = TRUE}.
diff --git a/R-package/R/generate.data.R b/R-package/R/generate.data.R
index fcbde084..a1a2c0d9 100644
--- a/R-package/R/generate.data.R
+++ b/R-package/R/generate.data.R
@@ -22,7 +22,7 @@
 #' where the covariance matrix has \eqn{(i,j)} entry equals \eqn{I(i \neq j)}.
 #' \code{cortype = 2} denotes the exponential structure,
 #' where the covariance matrix has \eqn{(i,j)} entry equals \eqn{rho^{|i-j|}}.
-#' code{cortype = 3} denotes the constant structure,
+#' \code{cortype = 3} denotes the constant structure,
 #' where the non-diagonal entries of covariance
 #' matrix are \eqn{rho} and diagonal entries are 1.
 #' @param snr A numerical value controlling the signal-to-noise ratio (SNR). The SNR is defined as
diff --git a/R-package/cran-comments.md b/R-package/cran-comments.md
index 6ab5f2b9..af5a2dc1 100644
--- a/R-package/cran-comments.md
+++ b/R-package/cran-comments.md
@@ -1,7 +1,3 @@
-## Comment
-
-`abess` is remove from CRAN because incorrect C++ flags in configure scripts. This submission has address this incorrectness.
-
 ## Test environments
 * local R installation, R 4.1.0
 * win-builder (devel)
diff --git a/R-package/man/abess.Rd b/R-package/man/abess.Rd
index 96a4143a..afe01dbf 100644
--- a/R-package/man/abess.Rd
+++ b/R-package/man/abess.Rd
@@ -257,10 +257,10 @@ Again, we suggest to use the default values unchanged because the same reason fo
 
 \code{abess} support some well-known advanced statistical methods to analyze data, including
 \itemize{
-\item{sure independent screening: } {helpful for ultra-high dimensional predictors (i.e., \eqn{p \gg n}). Use the parameter \code{screening.num} to retain the marginally most important predictors. See Fan et al (2008) for more details. }
-\item{best subset of group selection: } {helpful when predictors have group structure. Use the parameter \code{group.index} to specify the group structure of predictors. See Zhang et al (2021) for more details. }
-\item{\eqn{l_2} regularization best subset selection: } {helpful when signal-to-ratio is relatively small. Use the parameter \code{lambda} to control the magnitude of the regularization term.}
-\item{nuisance selection: } {helpful when the prior knowledge of important predictors is available. Use the parameter \code{always.include} to retain the important predictors.}
+\item sure independent screening: helpful for ultra-high dimensional predictors (i.e., \eqn{p \gg n}). Use the parameter \code{screening.num} to retain the marginally most important predictors. See Fan et al (2008) for more details.
+\item best subset of group selection: helpful when predictors have group structure. Use the parameter \code{group.index} to specify the group structure of predictors. See Zhang et al (2021) for more details.
+\item \eqn{l_2} regularization best subset selection: helpful when signal-to-ratio is relatively small. Use the parameter \code{lambda} to control the magnitude of the regularization term.
+\item nuisance selection: helpful when the prior knowledge of important predictors is available. Use the parameter \code{always.include} to retain the important predictors.
 }
 The arbitrary combination of the four methods are definitely support.
 Please see \href{https://abess-team.github.io/abess/articles/v07-advancedFeatures.html}{online vignettes} for more details about the advanced features support by \code{abess}.
diff --git a/R-package/man/coef.abesspca.Rd b/R-package/man/coef.abesspca.Rd
index a300f9fb..b8b58424 100644
--- a/R-package/man/coef.abesspca.Rd
+++ b/R-package/man/coef.abesspca.Rd
@@ -13,13 +13,13 @@
 the coefficient fitted at given \code{support.size}.
 If \code{support.size = NULL}, then all coefficients would be returned.
 Default: \code{support.size = NULL}.
-This parameter is omitted if {sparse.type = "kpc"}.}
+This parameter is omitted if \code{sparse.type = "kpc"}.}
 
 \item{kpc}{An integer vector specifies
 the coefficient fitted at given principal component.
 If \code{kpc = NULL}, then all coefficients would be returned.
 Default: \code{kpc = NULL}.
-This parameter is omitted if {sparse.type = "fpc"}.}
+This parameter is omitted if \code{sparse.type = "fpc"}.}
 
 \item{sparse}{A logical value, specifying whether the coefficients should be
 presented as sparse matrix or not. Default: \code{sparse = TRUE}.}
diff --git a/R-package/man/generate.data.Rd b/R-package/man/generate.data.Rd
index 1182ace9..19f0ee6b 100644
--- a/R-package/man/generate.data.Rd
+++ b/R-package/man/generate.data.Rd
@@ -48,7 +48,7 @@ If it is supplied, \code{support.size} would be omitted.}
 where the covariance matrix has \eqn{(i,j)} entry equals \eqn{I(i \neq j)}.
 \code{cortype = 2} denotes the exponential structure,
 where the covariance matrix has \eqn{(i,j)} entry equals \eqn{rho^{|i-j|}}.
-code{cortype = 3} denotes the constant structure,
+\code{cortype = 3} denotes the constant structure,
 where the non-diagonal entries of covariance
 matrix are \eqn{rho} and diagonal entries are 1.}